abbyswag/qwen-code
1
0
Fork 0
mirror of https://github.com/QwenLM/qwen-code.git synced 2025-12-20 08:47:44 +00:00
Code Issues Packages Projects Releases Wiki Activity

Merge remote-tracking branch 'upstream/main' into fix/issue-1186-schema-converter

Browse Source
This commit is contained in:
kefuxin
2025-12-16 11:08:51 +08:00
parent 44794121a8 8130020277
commit 573c33f68a
239 changed files with 9372 additions and 7082 deletions
Download Patch File Download Diff File Expand all files Collapse all files

13
docs/developers/tools/_meta.ts Normal file
View File

@@ -0,0 +1,13 @@
export default {
introduction: 'Introduction',
'file-system': 'File System',
'multi-file': 'Multi-File Read',
shell: 'Shell',
'todo-write': 'Todo Write',
task: 'Task',
'exit-plan-mode': 'Exit Plan Mode',
'web-fetch': 'Web Fetch',
'web-search': 'Web Search',
memory: 'Memory',
'mcp-server': 'MCP Servers',
};

149
docs/developers/tools/exit-plan-mode.md Normal file
View File

@@ -0,0 +1,149 @@
# Exit Plan Mode Tool (`exit_plan_mode`)
This document describes the `exit_plan_mode` tool for Qwen Code.
## Description
Use `exit_plan_mode` when you are in plan mode and have finished presenting your implementation plan. This tool prompts the user to approve or reject the plan and transitions from planning mode to implementation mode.
The tool is specifically designed for tasks that require planning implementation steps before writing code. It should NOT be used for research or information-gathering tasks.
### Arguments
`exit_plan_mode` takes one argument:
- `plan` (string, required): The implementation plan you want to present to the user for approval. This should be a concise, markdown-formatted plan describing the implementation steps.
## How to use `exit_plan_mode` with Qwen Code
The Exit Plan Mode tool is part of Qwen Code's planning workflow. When you're in plan mode (typically after exploring a codebase and designing an implementation approach), you use this tool to:
1. Present your implementation plan to the user
2. Request approval to proceed with implementation
3. Transition from plan mode to implementation mode based on user response
The tool will prompt the user with your plan and provide options to:
- **Proceed Once**: Approve the plan for this session only
- **Proceed Always**: Approve the plan and enable auto-approval for future edit operations
- **Cancel**: Reject the plan and remain in planning mode
Usage:
```
exit_plan_mode(plan="Your detailed implementation plan here...")
```
## When to Use This Tool
Use `exit_plan_mode` when:
1. **Implementation tasks**: You are planning the implementation steps for a coding task
2. **Plan completion**: You have finished exploring and designing your implementation approach
3. **User approval needed**: You need user confirmation before proceeding with code changes
4. **Code writing tasks**: The task involves writing, modifying, or refactoring code
### Examples of appropriate usage:
- "Help me implement user authentication" → Use after planning the auth system implementation
- "Add a new API endpoint for user management" → Use after designing the endpoint structure
- "Refactor the database layer to use TypeORM" → Use after planning the refactoring approach
## When NOT to Use This Tool
Do NOT use `exit_plan_mode` for:
1. **Research tasks**: Tasks focused on understanding or exploring existing code
2. **Information gathering**: When you're searching, reading, or analyzing without implementing
3. **Documentation tasks**: When creating documentation without code changes
4. **Analysis tasks**: When providing explanations or reviews without implementation
### Examples of inappropriate usage:
- "Search for and understand the implementation of vim mode in the codebase" → Don't use (research task)
- "Explain how the authentication system works" → Don't use (analysis task)
- "Find all TODO comments in the project" → Don't use (information gathering)
## `exit_plan_mode` examples
### Implementation planning example
```
exit_plan_mode(plan="## Implementation Plan for User Dashboard
### 1. Create Dashboard Components
- Create `UserDashboard.tsx` with main dashboard layout
- Add `DashboardStats.tsx` for displaying user statistics
- Implement `ActivityFeed.tsx` for recent user activities
### 2. Add API Integration
- Create `dashboardApi.ts` for dashboard data fetching
- Add endpoints for user stats and activity data
- Implement error handling and loading states
### 3. Update Routing
- Add dashboard route in `App.tsx`
- Update navigation component with dashboard link
- Add route protection for authenticated users
### 4. Testing and Documentation
- Write unit tests for dashboard components
- Add integration tests for API calls
- Update user documentation with dashboard features
This implementation will take approximately 4-6 hours and requires changes to 8-10 files.")
```
### Refactoring plan example
```
exit_plan_mode(plan="## Database Layer Refactoring Plan
### Phase 1: Setup and Configuration
- Install and configure TypeORM dependencies
- Create database connection configuration
- Set up entity definitions for existing models
### Phase 2: Migration
- Create migration scripts for existing data
- Implement new repository pattern using TypeORM
- Replace raw SQL queries with TypeORM query builder
### Phase 3: Testing and Validation
- Update all database tests to use TypeORM
- Validate data integrity after migration
- Performance testing to ensure no regressions
This refactoring will modernize our database layer while maintaining backward compatibility.")
```
## User Response Handling
After calling `exit_plan_mode`, the user can respond in several ways:
- **Proceed Once**: The plan is approved for immediate implementation with default confirmation settings
- **Proceed Always**: The plan is approved and auto-approval is enabled for subsequent edit operations
- **Cancel**: The plan is rejected, and the system remains in plan mode for further planning
The tool automatically adjusts the approval mode based on the user's choice, streamlining the implementation process according to user preferences.
## Important Notes
- **Plan mode only**: This tool should only be used when you are currently in plan mode
- **Implementation focus**: Only use for tasks that involve writing or modifying code
- **Concise plans**: Keep plans focused and concise - aim for clarity over exhaustive detail
- **Markdown support**: Plans support markdown formatting for better readability
- **Single use**: The tool should be used once per planning session when ready to proceed
- **User control**: The final decision to proceed always rests with the user
## Integration with Planning Workflow
The Exit Plan Mode tool is part of a larger planning workflow:
1. **Enter Plan Mode**: User requests or system determines planning is needed
2. **Exploration Phase**: Analyze codebase, understand requirements, explore options
3. **Plan Design**: Create implementation strategy based on exploration
4. **Plan Presentation**: Use `exit_plan_mode` to present plan to user
5. **Implementation Phase**: Upon approval, proceed with planned implementation
This workflow ensures thoughtful implementation approaches and gives users control over significant code changes.

167
docs/developers/tools/file-system.md Normal file
View File

@@ -0,0 +1,167 @@
# Qwen Code file system tools
Qwen Code provides a comprehensive suite of tools for interacting with the local file system. These tools allow the model to read from, write to, list, search, and modify files and directories, all under your control and typically with confirmation for sensitive operations.
**Note:** All file system tools operate within a `rootDirectory` (usually the current working directory where you launched the CLI) for security. Paths that you provide to these tools are generally expected to be absolute or are resolved relative to this root directory.
## 1. `list_directory` (ListFiles)
`list_directory` lists the names of files and subdirectories directly within a specified directory path. It can optionally ignore entries matching provided glob patterns.
- **Tool name:** `list_directory`
- **Display name:** ListFiles
- **File:** `ls.ts`
- **Parameters:**
- `path` (string, required): The absolute path to the directory to list.
- `ignore` (array of strings, optional): A list of glob patterns to exclude from the listing (e.g., `["*.log", ".git"]`).
- `respect_git_ignore` (boolean, optional): Whether to respect `.gitignore` patterns when listing files. Defaults to `true`.
- **Behavior:**
- Returns a list of file and directory names.
- Indicates whether each entry is a directory.
- Sorts entries with directories first, then alphabetically.
- **Output (`llmContent`):** A string like: `Directory listing for /path/to/your/folder:\n[DIR] subfolder1\nfile1.txt\nfile2.png`
- **Confirmation:** No.
## 2. `read_file` (ReadFile)
`read_file` reads and returns the content of a specified file. This tool handles text, images (PNG, JPG, GIF, WEBP, SVG, BMP), and PDF files. For text files, it can read specific line ranges. Other binary file types are generally skipped.
- **Tool name:** `read_file`
- **Display name:** ReadFile
- **File:** `read-file.ts`
- **Parameters:**
- `path` (string, required): The absolute path to the file to read.
- `offset` (number, optional): For text files, the 0-based line number to start reading from. Requires `limit` to be set.
- `limit` (number, optional): For text files, the maximum number of lines to read. If omitted, reads a default maximum (e.g., 2000 lines) or the entire file if feasible.
- **Behavior:**
- For text files: Returns the content. If `offset` and `limit` are used, returns only that slice of lines. Indicates if content was truncated due to line limits or line length limits.
- For image and PDF files: Returns the file content as a base64-encoded data structure suitable for model consumption.
- For other binary files: Attempts to identify and skip them, returning a message indicating it's a generic binary file.
- **Output:** (`llmContent`):
- For text files: The file content, potentially prefixed with a truncation message (e.g., `[File content truncated: showing lines 1-100 of 500 total lines...]\nActual file content...`).
- For image/PDF files: An object containing `inlineData` with `mimeType` and base64 `data` (e.g., `{ inlineData: { mimeType: 'image/png', data: 'base64encodedstring' } }`).
- For other binary files: A message like `Cannot display content of binary file: /path/to/data.bin`.
- **Confirmation:** No.
## 3. `write_file` (WriteFile)
`write_file` writes content to a specified file. If the file exists, it will be overwritten. If the file doesn't exist, it (and any necessary parent directories) will be created.
- **Tool name:** `write_file`
- **Display name:** WriteFile
- **File:** `write-file.ts`
- **Parameters:**
- `file_path` (string, required): The absolute path to the file to write to.
- `content` (string, required): The content to write into the file.
- **Behavior:**
- Writes the provided `content` to the `file_path`.
- Creates parent directories if they don't exist.
- **Output (`llmContent`):** A success message, e.g., `Successfully overwrote file: /path/to/your/file.txt` or `Successfully created and wrote to new file: /path/to/new/file.txt`.
- **Confirmation:** Yes. Shows a diff of changes and asks for user approval before writing.
## 4. `glob` (Glob)
`glob` finds files matching specific glob patterns (e.g., `src/**/*.ts`, `*.md`), returning absolute paths sorted by modification time (newest first).
- **Tool name:** `glob`
- **Display name:** Glob
- **File:** `glob.ts`
- **Parameters:**
- `pattern` (string, required): The glob pattern to match against (e.g., `"*.py"`, `"src/**/*.js"`).
- `path` (string, optional): The directory to search in. If not specified, the current working directory will be used.
- **Behavior:**
- Searches for files matching the glob pattern within the specified directory.
- Returns a list of absolute paths, sorted with the most recently modified files first.
- Respects .gitignore and .qwenignore patterns by default.
- Limits results to 100 files to prevent context overflow.
- **Output (`llmContent`):** A message like: `Found 5 file(s) matching "*.ts" within /path/to/search/dir, sorted by modification time (newest first):\n---\n/path/to/file1.ts\n/path/to/subdir/file2.ts\n---\n[95 files truncated] ...`
- **Confirmation:** No.
## 5. `grep_search` (Grep)
`grep_search` searches for a regular expression pattern within the content of files in a specified directory. Can filter files by a glob pattern. Returns the lines containing matches, along with their file paths and line numbers.
- **Tool name:** `grep_search`
- **Display name:** Grep
- **File:** `grep.ts` (with `ripGrep.ts` as fallback)
- **Parameters:**
- `pattern` (string, required): The regular expression pattern to search for in file contents (e.g., `"function\\s+myFunction"`, `"log.*Error"`).
- `path` (string, optional): File or directory to search in. Defaults to current working directory.
- `glob` (string, optional): Glob pattern to filter files (e.g. `"*.js"`, `"src/**/*.{ts,tsx}"`).
- `limit` (number, optional): Limit output to first N matching lines. Optional - shows all matches if not specified.
- **Behavior:**
- Uses ripgrep for fast search when available; otherwise falls back to a JavaScript-based search implementation.
- Returns matching lines with file paths and line numbers.
- Case-insensitive by default.
- Respects .gitignore and .qwenignore patterns.
- Limits output to prevent context overflow.
- **Output (`llmContent`):** A formatted string of matches, e.g.:
```
Found 3 matches for pattern "myFunction" in path "." (filter: "*.ts"):
---
src/utils.ts:15:export function myFunction() {
src/utils.ts:22: myFunction.call();
src/index.ts:5:import { myFunction } from './utils';
---
[0 lines truncated] ...
```
- **Confirmation:** No.
### `grep_search` examples
Search for a pattern with default result limiting:
```
grep_search(pattern="function\\s+myFunction", path="src")
```
Search for a pattern with custom result limiting:
```
grep_search(pattern="function", path="src", limit=50)
```
Search for a pattern with file filtering and custom result limiting:
```
grep_search(pattern="function", glob="*.js", limit=10)
```
## 6. `edit` (Edit)
`edit` replaces text within a file. By default it requires `old_string` to match a single unique location; set `replace_all` to `true` when you intentionally want to change every occurrence. This tool is designed for precise, targeted changes and requires significant context around the `old_string` to ensure it modifies the correct location.
- **Tool name:** `edit`
- **Display name:** Edit
- **File:** `edit.ts`
- **Parameters:**
- `file_path` (string, required): The absolute path to the file to modify.
- `old_string` (string, required): The exact literal text to replace.
**CRITICAL:** This string must uniquely identify the single instance to change. It should include sufficient context around the target text, matching whitespace and indentation precisely. If `old_string` is empty, the tool attempts to create a new file at `file_path` with `new_string` as content.
- `new_string` (string, required): The exact literal text to replace `old_string` with.
- `replace_all` (boolean, optional): Replace all occurrences of `old_string`. Defaults to `false`.
- **Behavior:**
- If `old_string` is empty and `file_path` does not exist, creates a new file with `new_string` as content.
- If `old_string` is provided, it reads the `file_path` and attempts to find exactly one occurrence unless `replace_all` is true.
- If the match is unique (or `replace_all` is true), it replaces the text with `new_string`.
- **Enhanced Reliability (Multi-Stage Edit Correction):** To significantly improve the success rate of edits, especially when the model-provided `old_string` might not be perfectly precise, the tool incorporates a multi-stage edit correction mechanism.
- If the initial `old_string` isn't found or matches multiple locations, the tool can leverage the Qwen model to iteratively refine `old_string` (and potentially `new_string`).
- This self-correction process attempts to identify the unique segment the model intended to modify, making the `edit` operation more robust even with slightly imperfect initial context.
- **Failure conditions:** Despite the correction mechanism, the tool will fail if:
- `file_path` is not absolute or is outside the root directory.
- `old_string` is not empty, but the `file_path` does not exist.
- `old_string` is empty, but the `file_path` already exists.
- `old_string` is not found in the file after attempts to correct it.
- `old_string` is found multiple times, `replace_all` is false, and the self-correction mechanism cannot resolve it to a single, unambiguous match.
- **Output (`llmContent`):**
- On success: `Successfully modified file: /path/to/file.txt (1 replacements).` or `Created new file: /path/to/new_file.txt with provided content.`
- On failure: An error message explaining the reason (e.g., `Failed to edit, 0 occurrences found...`, `Failed to edit because the text matches multiple locations...`).
- **Confirmation:** Yes. Shows a diff of the proposed changes and asks for user approval before writing to the file.
These file system tools provide a foundation for Qwen Code to understand and interact with your local project context.

62
docs/developers/tools/introduction.md Normal file
View File

@@ -0,0 +1,62 @@
# Qwen Code tools
Qwen Code includes built-in tools that the model uses to interact with your local environment, access information, and perform actions. These tools enhance the CLI's capabilities, enabling it to go beyond text generation and assist with a wide range of tasks.
## Overview of Qwen Code tools
In the context of Qwen Code, tools are specific functions or modules that the model can request to be executed. For example, if you ask the model to "Summarize the contents of `my_document.txt`," it will likely identify the need to read that file and will request the execution of the `read_file` tool.
The core component (`packages/core`) manages these tools, presents their definitions (schemas) to the model, executes them when requested, and returns the results to the model for further processing into a user-facing response.
These tools provide the following capabilities:
- **Access local information:** Tools allow the model to access your local file system, read file contents, list directories, etc.
- **Execute commands:** With tools like `run_shell_command`, the model can run shell commands (with appropriate safety measures and user confirmation).
- **Interact with the web:** Tools can fetch content from URLs.
- **Take actions:** Tools can modify files, write new files, or perform other actions on your system (again, typically with safeguards).
- **Ground responses:** By using tools to fetch real-time or specific local data, responses can be more accurate, relevant, and grounded in your actual context.
## How to use Qwen Code tools
To use Qwen Code tools, provide a prompt to the CLI. The process works as follows:
1. You provide a prompt to the CLI.
2. The CLI sends the prompt to the core.
3. The core, along with your prompt and conversation history, sends a list of available tools and their descriptions/schemas to the configured model API.
4. The model analyzes your request. If it determines that a tool is needed, its response will include a request to execute a specific tool with certain parameters.
5. The core receives this tool request, validates it, and (often after user confirmation for sensitive operations) executes the tool.
6. The output from the tool is sent back to the model.
7. The model uses the tool's output to formulate its final answer, which is then sent back through the core to the CLI and displayed to you.
You will typically see messages in the CLI indicating when a tool is being called and whether it succeeded or failed.
## Security and confirmation
Many tools, especially those that can modify your file system or execute commands (`write_file`, `edit`, `run_shell_command`), are designed with safety in mind. Qwen Code will typically:
- **Require confirmation:** Prompt you before executing potentially sensitive operations, showing you what action is about to be taken.
- **Utilize sandboxing:** All tools are subject to restrictions enforced by sandboxing (see [Sandboxing in Qwen Code](../sandbox.md)). This means that when operating in a sandbox, any tools (including MCP servers) you wish to use must be available _inside_ the sandbox environment. For example, to run an MCP server through `npx`, the `npx` executable must be installed within the sandbox's Docker image or be available in the `sandbox-exec` environment.
It's important to always review confirmation prompts carefully before allowing a tool to proceed.
## Learn more about Qwen Code's tools
Qwen Code's built-in tools can be broadly categorized as follows:
- **[File System Tools](./file-system.md):** For interacting with files and directories (reading, writing, listing, searching, etc.).
- **[Shell Tool](./shell.md) (`run_shell_command`):** For executing shell commands.
- **[Web Fetch Tool](./web-fetch.md) (`web_fetch`):** For retrieving content from URLs.
- **[Web Search Tool](./web-search.md) (`web_search`):** For searching the web.
- **[Multi-File Read Tool](./multi-file.md) (`read_many_files`):** A specialized tool for reading content from multiple files or directories, often used by the `@` command.
- **[Memory Tool](./memory.md) (`save_memory`):** For saving and recalling information across sessions.
- **[Todo Write Tool](./todo-write.md) (`todo_write`):** For creating and managing structured task lists during coding sessions.
- **[Task Tool](./task.md) (`task`):** For delegating complex tasks to specialized subagents.
- **[Exit Plan Mode Tool](./exit-plan-mode.md) (`exit_plan_mode`):** For exiting plan mode and proceeding with implementation.
Additionally, these tools incorporate:
- **[MCP servers](./mcp-server.md)**: MCP servers act as a bridge between the model and your local environment or other services like APIs.
- **[MCP Quick Start Guide](../mcp-quick-start.md)**: Get started with MCP in 5 minutes with practical examples
- **[MCP Example Configurations](../mcp-example-configs.md)**: Ready-to-use configurations for common scenarios
- **[MCP Testing & Validation](../mcp-testing-validation.md)**: Test and validate your MCP server setups
- **[Sandboxing](../sandbox.md)**: Sandboxing isolates the model and its changes from your environment to reduce potential risk.

871
docs/developers/tools/mcp-server.md Normal file
View File

@@ -0,0 +1,871 @@
# MCP servers with Qwen Code
This document provides a guide to configuring and using Model Context Protocol (MCP) servers with Qwen Code.
## What is an MCP server?
An MCP server is an application that exposes tools and resources to the CLI through the Model Context Protocol, allowing it to interact with external systems and data sources. MCP servers act as a bridge between the model and your local environment or other services like APIs.
An MCP server enables the CLI to:
- **Discover tools:** List available tools, their descriptions, and parameters through standardized schema definitions.
- **Execute tools:** Call specific tools with defined arguments and receive structured responses.
- **Access resources:** Read data from specific resources (though the CLI primarily focuses on tool execution).
With an MCP server, you can extend the CLI's capabilities to perform actions beyond its built-in features, such as interacting with databases, APIs, custom scripts, or specialized workflows.
## Core Integration Architecture
Qwen Code integrates with MCP servers through a sophisticated discovery and execution system built into the core package (`packages/core/src/tools/`):
### Discovery Layer (`mcp-client.ts`)
The discovery process is orchestrated by `discoverMcpTools()`, which:
1. **Iterates through configured servers** from your `settings.json` `mcpServers` configuration
2. **Establishes connections** using appropriate transport mechanisms (Stdio, SSE, or Streamable HTTP)
3. **Fetches tool definitions** from each server using the MCP protocol
4. **Sanitizes and validates** tool schemas for compatibility with the Qwen API
5. **Registers tools** in the global tool registry with conflict resolution
### Execution Layer (`mcp-tool.ts`)
Each discovered MCP tool is wrapped in a `DiscoveredMCPTool` instance that:
- **Handles confirmation logic** based on server trust settings and user preferences
- **Manages tool execution** by calling the MCP server with proper parameters
- **Processes responses** for both the LLM context and user display
- **Maintains connection state** and handles timeouts
### Transport Mechanisms
The CLI supports three MCP transport types:
- **Stdio Transport:** Spawns a subprocess and communicates via stdin/stdout
- **SSE Transport:** Connects to Server-Sent Events endpoints
- **Streamable HTTP Transport:** Uses HTTP streaming for communication
## How to set up your MCP server
Qwen Code uses the `mcpServers` configuration in your `settings.json` file to locate and connect to MCP servers. This configuration supports multiple servers with different transport mechanisms.
### Configure the MCP server in settings.json
You can configure MCP servers in your `settings.json` file in two main ways: through the top-level `mcpServers` object for specific server definitions, and through the `mcp` object for global settings that control server discovery and execution.
#### Global MCP Settings (`mcp`)
The `mcp` object in your `settings.json` allows you to define global rules for all MCP servers.
- **`mcp.serverCommand`** (string): A global command to start an MCP server.
- **`mcp.allowed`** (array of strings): A list of MCP server names to allow. If this is set, only servers from this list (matching the keys in the `mcpServers` object) will be connected to.
- **`mcp.excluded`** (array of strings): A list of MCP server names to exclude. Servers in this list will not be connected to.
**Example:**
```json
{
"mcp": {
"allowed": ["my-trusted-server"],
"excluded": ["experimental-server"]
}
}
```
#### Server-Specific Configuration (`mcpServers`)
The `mcpServers` object is where you define each individual MCP server you want the CLI to connect to.
### Configuration Structure
Add an `mcpServers` object to your `settings.json` file:
```json
{ ...file contains other config objects
"mcpServers": {
"serverName": {
"command": "path/to/server",
"args": ["--arg1", "value1"],
"env": {
"API_KEY": "$MY_API_TOKEN"
},
"cwd": "./server-directory",
"timeout": 30000,
"trust": false
}
}
}
```
### Configuration Properties
Each server configuration supports the following properties:
#### Required (one of the following)
- **`command`** (string): Path to the executable for Stdio transport
- **`url`** (string): SSE endpoint URL (e.g., `"http://localhost:8080/sse"`)
- **`httpUrl`** (string): HTTP streaming endpoint URL
#### Optional
- **`args`** (string[]): Command-line arguments for Stdio transport
- **`headers`** (object): Custom HTTP headers when using `url` or `httpUrl`
- **`env`** (object): Environment variables for the server process. Values can reference environment variables using `$VAR_NAME` or `${VAR_NAME}` syntax
- **`cwd`** (string): Working directory for Stdio transport
- **`timeout`** (number): Request timeout in milliseconds (default: 600,000ms = 10 minutes)
- **`trust`** (boolean): When `true`, bypasses all tool call confirmations for this server (default: `false`)
- **`includeTools`** (string[]): List of tool names to include from this MCP server. When specified, only the tools listed here will be available from this server (allowlist behavior). If not specified, all tools from the server are enabled by default.
- **`excludeTools`** (string[]): List of tool names to exclude from this MCP server. Tools listed here will not be available to the model, even if they are exposed by the server. **Note:** `excludeTools` takes precedence over `includeTools` - if a tool is in both lists, it will be excluded.
- **`targetAudience`** (string): The OAuth Client ID allowlisted on the IAP-protected application you are trying to access. Used with `authProviderType: 'service_account_impersonation'`.
- **`targetServiceAccount`** (string): The email address of the Google Cloud Service Account to impersonate. Used with `authProviderType: 'service_account_impersonation'`.
### OAuth Support for Remote MCP Servers
Qwen Code supports OAuth 2.0 authentication for remote MCP servers using SSE or HTTP transports. This enables secure access to MCP servers that require authentication.
#### Automatic OAuth Discovery
For servers that support OAuth discovery, you can omit the OAuth configuration and let the CLI discover it automatically:
```json
{
"mcpServers": {
"discoveredServer": {
"url": "https://api.example.com/sse"
}
}
}
```
The CLI will automatically:
- Detect when a server requires OAuth authentication (401 responses)
- Discover OAuth endpoints from server metadata
- Perform dynamic client registration if supported
- Handle the OAuth flow and token management
#### Authentication Flow
When connecting to an OAuth-enabled server:
1. **Initial connection attempt** fails with 401 Unauthorized
2. **OAuth discovery** finds authorization and token endpoints
3. **Browser opens** for user authentication (requires local browser access)
4. **Authorization code** is exchanged for access tokens
5. **Tokens are stored** securely for future use
6. **Connection retry** succeeds with valid tokens
#### Browser Redirect Requirements
**Important:** OAuth authentication requires that your local machine can:
- Open a web browser for authentication
- Receive redirects on `http://localhost:7777/oauth/callback`
This feature will not work in:
- Headless environments without browser access
- Remote SSH sessions without X11 forwarding
- Containerized environments without browser support
#### Managing OAuth Authentication
Use the `/mcp auth` command to manage OAuth authentication:
```bash
# List servers requiring authentication
/mcp auth
# Authenticate with a specific server
/mcp auth serverName
# Re-authenticate if tokens expire
/mcp auth serverName
```
#### OAuth Configuration Properties
- **`enabled`** (boolean): Enable OAuth for this server
- **`clientId`** (string): OAuth client identifier (optional with dynamic registration)
- **`clientSecret`** (string): OAuth client secret (optional for public clients)
- **`authorizationUrl`** (string): OAuth authorization endpoint (auto-discovered if omitted)
- **`tokenUrl`** (string): OAuth token endpoint (auto-discovered if omitted)
- **`scopes`** (string[]): Required OAuth scopes
- **`redirectUri`** (string): Custom redirect URI (defaults to `http://localhost:7777/oauth/callback`)
- **`tokenParamName`** (string): Query parameter name for tokens in SSE URLs
- **`audiences`** (string[]): Audiences the token is valid for
#### Token Management
OAuth tokens are automatically:
- **Stored securely** in `~/.qwen/mcp-oauth-tokens.json`
- **Refreshed** when expired (if refresh tokens are available)
- **Validated** before each connection attempt
- **Cleaned up** when invalid or expired
#### Authentication Provider Type
You can specify the authentication provider type using the `authProviderType` property:
- **`authProviderType`** (string): Specifies the authentication provider. Can be one of the following:
- **`dynamic_discovery`** (default): The CLI will automatically discover the OAuth configuration from the server.
- **`google_credentials`**: The CLI will use the Google Application Default Credentials (ADC) to authenticate with the server. When using this provider, you must specify the required scopes.
- **`service_account_impersonation`**: The CLI will impersonate a Google Cloud Service Account to authenticate with the server. This is useful for accessing IAP-protected services (this was specifically designed for Cloud Run services).
#### Google Credentials
```json
{
"mcpServers": {
"googleCloudServer": {
"httpUrl": "https://my-gcp-service.run.app/mcp",
"authProviderType": "google_credentials",
"oauth": {
"scopes": ["https://www.googleapis.com/auth/userinfo.email"]
}
}
}
}
```
#### Service Account Impersonation
To authenticate with a server using Service Account Impersonation, you must set the `authProviderType` to `service_account_impersonation` and provide the following properties:
- **`targetAudience`** (string): The OAuth Client ID allowslisted on the IAP-protected application you are trying to access.
- **`targetServiceAccount`** (string): The email address of the Google Cloud Service Account to impersonate.
The CLI will use your local Application Default Credentials (ADC) to generate an OIDC ID token for the specified service account and audience. This token will then be used to authenticate with the MCP server.
#### Setup Instructions
1. **[Create](https://cloud.google.com/iap/docs/oauth-client-creation) or use an existing OAuth 2.0 client ID.** To use an existing OAuth 2.0 client ID, follow the steps in [How to share OAuth Clients](https://cloud.google.com/iap/docs/sharing-oauth-clients).
2. **Add the OAuth ID to the allowlist for [programmatic access](https://cloud.google.com/iap/docs/sharing-oauth-clients#programmatic_access) for the application.** Since Cloud Run is not yet a supported resource type in gcloud iap, you must allowlist the Client ID on the project.
3. **Create a service account.** [Documentation](https://cloud.google.com/iam/docs/service-accounts-create#creating), [Cloud Console Link](https://console.cloud.google.com/iam-admin/serviceaccounts)
4. **Add both the service account and users to the IAP Policy** in the "Security" tab of the Cloud Run service itself or via gcloud.
5. **Grant all users and groups** who will access the MCP Server the necessary permissions to [impersonate the service account](https://cloud.google.com/docs/authentication/use-service-account-impersonation) (i.e., `roles/iam.serviceAccountTokenCreator`).
6. **[Enable](https://console.cloud.google.com/apis/library/iamcredentials.googleapis.com) the IAM Credentials API** for your project.
### Example Configurations
#### Python MCP Server (Stdio)
```json
{
"mcpServers": {
"pythonTools": {
"command": "python",
"args": ["-m", "my_mcp_server", "--port", "8080"],
"cwd": "./mcp-servers/python",
"env": {
"DATABASE_URL": "$DB_CONNECTION_STRING",
"API_KEY": "${EXTERNAL_API_KEY}"
},
"timeout": 15000
}
}
}
```
#### Node.js MCP Server (Stdio)
```json
{
"mcpServers": {
"nodeServer": {
"command": "node",
"args": ["dist/server.js", "--verbose"],
"cwd": "./mcp-servers/node",
"trust": true
}
}
}
```
#### Docker-based MCP Server
```json
{
"mcpServers": {
"dockerizedServer": {
"command": "docker",
"args": [
"run",
"-i",
"--rm",
"-e",
"API_KEY",
"-v",
"${PWD}:/workspace",
"my-mcp-server:latest"
],
"env": {
"API_KEY": "$EXTERNAL_SERVICE_TOKEN"
}
}
}
}
```
#### HTTP-based MCP Server
```json
{
"mcpServers": {
"httpServer": {
"httpUrl": "http://localhost:3000/mcp",
"timeout": 5000
}
}
}
```
#### HTTP-based MCP Server with Custom Headers
```json
{
"mcpServers": {
"httpServerWithAuth": {
"httpUrl": "http://localhost:3000/mcp",
"headers": {
"Authorization": "Bearer your-api-token",
"X-Custom-Header": "custom-value",
"Content-Type": "application/json"
},
"timeout": 5000
}
}
}
```
#### MCP Server with Tool Filtering
```json
{
"mcpServers": {
"filteredServer": {
"command": "python",
"args": ["-m", "my_mcp_server"],
"includeTools": ["safe_tool", "file_reader", "data_processor"],
// "excludeTools": ["dangerous_tool", "file_deleter"],
"timeout": 30000
}
}
}
```
### SSE MCP Server with SA Impersonation
```json
{
"mcpServers": {
"myIapProtectedServer": {
"url": "https://my-iap-service.run.app/sse",
"authProviderType": "service_account_impersonation",
"targetAudience": "YOUR_IAP_CLIENT_ID.apps.googleusercontent.com",
"targetServiceAccount": "your-sa@your-project.iam.gserviceaccount.com"
}
}
}
```
## Discovery Process Deep Dive
When Qwen Code starts, it performs MCP server discovery through the following detailed process:
### 1. Server Iteration and Connection
For each configured server in `mcpServers`:
1. **Status tracking begins:** Server status is set to `CONNECTING`
2. **Transport selection:** Based on configuration properties:
- `httpUrl` → `StreamableHTTPClientTransport`
- `url` → `SSEClientTransport`
- `command` → `StdioClientTransport`
3. **Connection establishment:** The MCP client attempts to connect with the configured timeout
4. **Error handling:** Connection failures are logged and the server status is set to `DISCONNECTED`
### 2. Tool Discovery
Upon successful connection:
1. **Tool listing:** The client calls the MCP server's tool listing endpoint
2. **Schema validation:** Each tool's function declaration is validated
3. **Tool filtering:** Tools are filtered based on `includeTools` and `excludeTools` configuration
4. **Name sanitization:** Tool names are cleaned to meet Qwen API requirements:
- Invalid characters (non-alphanumeric, underscore, dot, hyphen) are replaced with underscores
- Names longer than 63 characters are truncated with middle replacement (`___`)
### 3. Conflict Resolution
When multiple servers expose tools with the same name:
1. **First registration wins:** The first server to register a tool name gets the unprefixed name
2. **Automatic prefixing:** Subsequent servers get prefixed names: `serverName__toolName`
3. **Registry tracking:** The tool registry maintains mappings between server names and their tools
### 4. Schema Processing
Tool parameter schemas undergo sanitization for API compatibility:
- **`$schema` properties** are removed
- **`additionalProperties`** are stripped
- **`anyOf` with `default`** have their default values removed (Vertex AI compatibility)
- **Recursive processing** applies to nested schemas
### 5. Connection Management
After discovery:
- **Persistent connections:** Servers that successfully register tools maintain their connections
- **Cleanup:** Servers that provide no usable tools have their connections closed
- **Status updates:** Final server statuses are set to `CONNECTED` or `DISCONNECTED`
## Tool Execution Flow
When the model decides to use an MCP tool, the following execution flow occurs:
### 1. Tool Invocation
The model generates a `FunctionCall` with:
- **Tool name:** The registered name (potentially prefixed)
- **Arguments:** JSON object matching the tool's parameter schema
### 2. Confirmation Process
Each `DiscoveredMCPTool` implements sophisticated confirmation logic:
#### Trust-based Bypass
```typescript
if (this.trust) {
return false; // No confirmation needed
}
```
#### Dynamic Allow-listing
The system maintains internal allow-lists for:
- **Server-level:** `serverName` → All tools from this server are trusted
- **Tool-level:** `serverName.toolName` → This specific tool is trusted
#### User Choice Handling
When confirmation is required, users can choose:
- **Proceed once:** Execute this time only
- **Always allow this tool:** Add to tool-level allow-list
- **Always allow this server:** Add to server-level allow-list
- **Cancel:** Abort execution
### 3. Execution
Upon confirmation (or trust bypass):
1. **Parameter preparation:** Arguments are validated against the tool's schema
2. **MCP call:** The underlying `CallableTool` invokes the server with:
```typescript
const functionCalls = [
{
name: this.serverToolName, // Original server tool name
args: params,
},
];
```
3. **Response processing:** Results are formatted for both LLM context and user display
### 4. Response Handling
The execution result contains:
- **`llmContent`:** Raw response parts for the language model's context
- **`returnDisplay`:** Formatted output for user display (often JSON in markdown code blocks)
## How to interact with your MCP server
### Using the `/mcp` Command
The `/mcp` command provides comprehensive information about your MCP server setup:
```bash
/mcp
```
This displays:
- **Server list:** All configured MCP servers
- **Connection status:** `CONNECTED`, `CONNECTING`, or `DISCONNECTED`
- **Server details:** Configuration summary (excluding sensitive data)
- **Available tools:** List of tools from each server with descriptions
- **Discovery state:** Overall discovery process status
### Example `/mcp` Output
```
MCP Servers Status:
📡 pythonTools (CONNECTED)
Command: python -m my_mcp_server --port 8080
Working Directory: ./mcp-servers/python
Timeout: 15000ms
Tools: calculate_sum, file_analyzer, data_processor
🔌 nodeServer (DISCONNECTED)
Command: node dist/server.js --verbose
Error: Connection refused
🐳 dockerizedServer (CONNECTED)
Command: docker run -i --rm -e API_KEY my-mcp-server:latest
Tools: docker__deploy, docker__status
Discovery State: COMPLETED
```
### Tool Usage
Once discovered, MCP tools are available to the Qwen model like built-in tools. The model will automatically:
1. **Select appropriate tools** based on your requests
2. **Present confirmation dialogs** (unless the server is trusted)
3. **Execute tools** with proper parameters
4. **Display results** in a user-friendly format
## Status Monitoring and Troubleshooting
### Connection States
The MCP integration tracks several states:
#### Server Status (`MCPServerStatus`)
- **`DISCONNECTED`:** Server is not connected or has errors
- **`CONNECTING`:** Connection attempt in progress
- **`CONNECTED`:** Server is connected and ready
#### Discovery State (`MCPDiscoveryState`)
- **`NOT_STARTED`:** Discovery hasn't begun
- **`IN_PROGRESS`:** Currently discovering servers
- **`COMPLETED`:** Discovery finished (with or without errors)
### Common Issues and Solutions
#### Server Won't Connect
**Symptoms:** Server shows `DISCONNECTED` status
**Troubleshooting:**
1. **Check configuration:** Verify `command`, `args`, and `cwd` are correct
2. **Test manually:** Run the server command directly to ensure it works
3. **Check dependencies:** Ensure all required packages are installed
4. **Review logs:** Look for error messages in the CLI output
5. **Verify permissions:** Ensure the CLI can execute the server command
#### No Tools Discovered
**Symptoms:** Server connects but no tools are available
**Troubleshooting:**
1. **Verify tool registration:** Ensure your server actually registers tools
2. **Check MCP protocol:** Confirm your server implements the MCP tool listing correctly
3. **Review server logs:** Check stderr output for server-side errors
4. **Test tool listing:** Manually test your server's tool discovery endpoint
#### Tools Not Executing
**Symptoms:** Tools are discovered but fail during execution
**Troubleshooting:**
1. **Parameter validation:** Ensure your tool accepts the expected parameters
2. **Schema compatibility:** Verify your input schemas are valid JSON Schema
3. **Error handling:** Check if your tool is throwing unhandled exceptions
4. **Timeout issues:** Consider increasing the `timeout` setting
#### Sandbox Compatibility
**Symptoms:** MCP servers fail when sandboxing is enabled
**Solutions:**
1. **Docker-based servers:** Use Docker containers that include all dependencies
2. **Path accessibility:** Ensure server executables are available in the sandbox
3. **Network access:** Configure sandbox to allow necessary network connections
4. **Environment variables:** Verify required environment variables are passed through
### Debugging Tips
1. **Enable debug mode:** Run the CLI with `--debug` for verbose output
2. **Check stderr:** MCP server stderr is captured and logged (INFO messages filtered)
3. **Test isolation:** Test your MCP server independently before integrating
4. **Incremental setup:** Start with simple tools before adding complex functionality
5. **Use `/mcp` frequently:** Monitor server status during development
## Important Notes
### Security Considerations
- **Trust settings:** The `trust` option bypasses all confirmation dialogs. Use cautiously and only for servers you completely control
- **Access tokens:** Be security-aware when configuring environment variables containing API keys or tokens
- **Sandbox compatibility:** When using sandboxing, ensure MCP servers are available within the sandbox environment
- **Private data:** Using broadly scoped personal access tokens can lead to information leakage between repositories
### Performance and Resource Management
- **Connection persistence:** The CLI maintains persistent connections to servers that successfully register tools
- **Automatic cleanup:** Connections to servers providing no tools are automatically closed
- **Timeout management:** Configure appropriate timeouts based on your server's response characteristics
- **Resource monitoring:** MCP servers run as separate processes and consume system resources
### Schema Compatibility
- **Schema compliance mode:** By default (`schemaCompliance: "auto"`), tool schemas are passed through as-is. Set `"model": { "generationConfig": { "schemaCompliance": "openapi_30" } }` in your `settings.json` to convert models to Strict OpenAPI 3.0 format.
- **OpenAPI 3.0 transformations:** When `openapi_30` mode is enabled, the system handles:
- Nullable types: `["string", "null"]` -> `type: "string", nullable: true`
- Const values: `const: "foo"` -> `enum: ["foo"]`
- Exclusive limits: numeric `exclusiveMinimum` -> boolean form with `minimum`
- Keyword removal: `$schema`, `$id`, `dependencies`, `patternProperties`
- **Name sanitization:** Tool names are automatically sanitized to meet API requirements
- **Conflict resolution:** Tool name conflicts between servers are resolved through automatic prefixing
This comprehensive integration makes MCP servers a powerful way to extend the CLI's capabilities while maintaining security, reliability, and ease of use.
## Returning Rich Content from Tools
MCP tools are not limited to returning simple text. You can return rich, multi-part content, including text, images, audio, and other binary data in a single tool response. This allows you to build powerful tools that can provide diverse information to the model in a single turn.
All data returned from the tool is processed and sent to the model as context for its next generation, enabling it to reason about or summarize the provided information.
### How It Works
To return rich content, your tool's response must adhere to the MCP specification for a [`CallToolResult`](https://modelcontextprotocol.io/specification/2025-06-18/server/tools#tool-result). The `content` field of the result should be an array of `ContentBlock` objects. The CLI will correctly process this array, separating text from binary data and packaging it for the model.
You can mix and match different content block types in the `content` array. The supported block types include:
- `text`
- `image`
- `audio`
- `resource` (embedded content)
- `resource_link`
### Example: Returning Text and an Image
Here is an example of a valid JSON response from an MCP tool that returns both a text description and an image:
```json
{
"content": [
{
"type": "text",
"text": "Here is the logo you requested."
},
{
"type": "image",
"data": "BASE64_ENCODED_IMAGE_DATA_HERE",
"mimeType": "image/png"
},
{
"type": "text",
"text": "The logo was created in 2025."
}
]
}
```
When Qwen Code receives this response, it will:
1. Extract all the text and combine it into a single `functionResponse` part for the model.
2. Present the image data as a separate `inlineData` part.
3. Provide a clean, user-friendly summary in the CLI, indicating that both text and an image were received.
This enables you to build sophisticated tools that can provide rich, multi-modal context to the Qwen model.
## MCP Prompts as Slash Commands
In addition to tools, MCP servers can expose predefined prompts that can be executed as slash commands within Qwen Code. This allows you to create shortcuts for common or complex queries that can be easily invoked by name.
### Defining Prompts on the Server
Here's a small example of a stdio MCP server that defines prompts:
```ts
import { McpServer } from '@modelcontextprotocol/sdk/server/mcp.js';
import { StdioServerTransport } from '@modelcontextprotocol/sdk/server/stdio.js';
import { z } from 'zod';
const server = new McpServer({
name: 'prompt-server',
version: '1.0.0',
});
server.registerPrompt(
'poem-writer',
{
title: 'Poem Writer',
description: 'Write a nice haiku',
argsSchema: { title: z.string(), mood: z.string().optional() },
},
({ title, mood }) => ({
messages: [
{
role: 'user',
content: {
type: 'text',
text: `Write a haiku${mood ? ` with the mood ${mood}` : ''} called ${title}. Note that a haiku is 5 syllables followed by 7 syllables followed by 5 syllables `,
},
},
],
}),
);
const transport = new StdioServerTransport();
await server.connect(transport);
```
This can be included in `settings.json` under `mcpServers` with:
```json
{
"mcpServers": {
"nodeServer": {
"command": "node",
"args": ["filename.ts"]
}
}
}
```
### Invoking Prompts
Once a prompt is discovered, you can invoke it using its name as a slash command. The CLI will automatically handle parsing arguments.
```bash
/poem-writer --title="Qwen Code" --mood="reverent"
```
or, using positional arguments:
```bash
/poem-writer "Qwen Code" reverent
```
When you run this command, the CLI executes the `prompts/get` method on the MCP server with the provided arguments. The server is responsible for substituting the arguments into the prompt template and returning the final prompt text. The CLI then sends this prompt to the model for execution. This provides a convenient way to automate and share common workflows.
## Managing MCP Servers with `qwen mcp`
While you can always configure MCP servers by manually editing your `settings.json` file, the CLI provides a convenient set of commands to manage your server configurations programmatically. These commands streamline the process of adding, listing, and removing MCP servers without needing to directly edit JSON files.
### Adding a Server (`qwen mcp add`)
The `add` command configures a new MCP server in your `settings.json`. Based on the scope (`-s, --scope`), it will be added to either the user config `~/.qwen/settings.json` or the project config `.qwen/settings.json` file.
**Command:**
```bash
qwen mcp add [options] <name> <commandOrUrl> [args...]
```
- `<name>`: A unique name for the server.
- `<commandOrUrl>`: The command to execute (for `stdio`) or the URL (for `http`/`sse`).
- `[args...]`: Optional arguments for a `stdio` command.
**Options (Flags):**
- `-s, --scope`: Configuration scope (user or project). [default: "project"]
- `-t, --transport`: Transport type (stdio, sse, http). [default: "stdio"]
- `-e, --env`: Set environment variables (e.g. -e KEY=value).
- `-H, --header`: Set HTTP headers for SSE and HTTP transports (e.g. -H "X-Api-Key: abc123" -H "Authorization: Bearer abc123").
- `--timeout`: Set connection timeout in milliseconds.
- `--trust`: Trust the server (bypass all tool call confirmation prompts).
- `--description`: Set the description for the server.
- `--include-tools`: A comma-separated list of tools to include.
- `--exclude-tools`: A comma-separated list of tools to exclude.
#### Adding an stdio server
This is the default transport for running local servers.
```bash
# Basic syntax
qwen mcp add <name> <command> [args...]
# Example: Adding a local server
qwen mcp add my-stdio-server -e API_KEY=123 /path/to/server arg1 arg2 arg3
# Example: Adding a local python server
qwen mcp add python-server python server.py --port 8080
```
#### Adding an HTTP server
This transport is for servers that use the streamable HTTP transport.
```bash
# Basic syntax
qwen mcp add --transport http <name> <url>
# Example: Adding an HTTP server
qwen mcp add --transport http http-server https://api.example.com/mcp/
# Example: Adding an HTTP server with an authentication header
qwen mcp add --transport http secure-http https://api.example.com/mcp/ --header "Authorization: Bearer abc123"
```
#### Adding an SSE server
This transport is for servers that use Server-Sent Events (SSE).
```bash
# Basic syntax
qwen mcp add --transport sse <name> <url>
# Example: Adding an SSE server
qwen mcp add --transport sse sse-server https://api.example.com/sse/
# Example: Adding an SSE server with an authentication header
qwen mcp add --transport sse secure-sse https://api.example.com/sse/ --header "Authorization: Bearer abc123"
```
### Listing Servers (`qwen mcp list`)
To view all MCP servers currently configured, use the `list` command. It displays each server's name, configuration details, and connection status.
**Command:**
```bash
qwen mcp list
```
**Example Output:**
```sh
✓ stdio-server: command: python3 server.py (stdio) - Connected
✓ http-server: https://api.example.com/mcp (http) - Connected
✗ sse-server: https://api.example.com/sse (sse) - Disconnected
```
### Removing a Server (`qwen mcp remove`)
To delete a server from your configuration, use the `remove` command with the server's name.
**Command:**
```bash
qwen mcp remove <name>
```
**Example:**
```bash
qwen mcp remove my-server
```
This will find and delete the "my-server" entry from the `mcpServers` object in the appropriate `settings.json` file based on the scope (`-s, --scope`).

44
docs/developers/tools/memory.md Normal file
View File

@@ -0,0 +1,44 @@
# Memory Tool (`save_memory`)
This document describes the `save_memory` tool for Qwen Code.
## Description
Use `save_memory` to save and recall information across your Qwen Code sessions. With `save_memory`, you can direct the CLI to remember key details across sessions, providing personalized and directed assistance.
### Arguments
`save_memory` takes one argument:
- `fact` (string, required): The specific fact or piece of information to remember. This should be a clear, self-contained statement written in natural language.
## How to use `save_memory` with Qwen Code
The tool appends the provided `fact` to your context file in the user's home directory (`~/.qwen/QWEN.md` by default). This filename can be configured via `contextFileName`.
Once added, the facts are stored under a `## Qwen Added Memories` section. This file is loaded as context in subsequent sessions, allowing the CLI to recall the saved information.
Usage:
```
save_memory(fact="Your fact here.")
```
### `save_memory` examples
Remember a user preference:
```
save_memory(fact="My preferred programming language is Python.")
```
Store a project-specific detail:
```
save_memory(fact="The project I'm currently working on is called 'qwen-code'.")
```
## Important notes
- **General usage:** This tool should be used for concise, important facts. It is not intended for storing large amounts of data or conversational history.
- **Memory file:** The memory file is a plain text Markdown file, so you can view and edit it manually if needed.

69
docs/developers/tools/multi-file.md Normal file
View File

@@ -0,0 +1,69 @@
# Multi File Read Tool (`read_many_files`)
This document describes the `read_many_files` tool for Qwen Code.
## Description
Use `read_many_files` to read content from multiple files specified by paths or glob patterns. The behavior of this tool depends on the provided files:
- For text files, this tool concatenates their content into a single string.
- For image (e.g., PNG, JPEG), PDF, audio (MP3, WAV), and video (MP4, MOV) files, it reads and returns them as base64-encoded data, provided they are explicitly requested by name or extension.
`read_many_files` can be used to perform tasks such as getting an overview of a codebase, finding where specific functionality is implemented, reviewing documentation, or gathering context from multiple configuration files.
**Note:** `read_many_files` looks for files following the provided paths or glob patterns. A directory path such as `"/docs"` will return an empty result; the tool requires a pattern such as `"/docs/*"` or `"/docs/*.md"` to identify the relevant files.
### Arguments
`read_many_files` takes the following arguments:
- `paths` (list[string], required): An array of glob patterns or paths relative to the tool's target directory (e.g., `["src/**/*.ts"]`, `["README.md", "docs/*", "assets/logo.png"]`).
- `exclude` (list[string], optional): Glob patterns for files/directories to exclude (e.g., `["**/*.log", "temp/"]`). These are added to default excludes if `useDefaultExcludes` is true.
- `include` (list[string], optional): Additional glob patterns to include. These are merged with `paths` (e.g., `["*.test.ts"]` to specifically add test files if they were broadly excluded, or `["images/*.jpg"]` to include specific image types).
- `recursive` (boolean, optional): Whether to search recursively. This is primarily controlled by `**` in glob patterns. Defaults to `true`.
- `useDefaultExcludes` (boolean, optional): Whether to apply a list of default exclusion patterns (e.g., `node_modules`, `.git`, non image/pdf binary files). Defaults to `true`.
- `respect_git_ignore` (boolean, optional): Whether to respect .gitignore patterns when finding files. Defaults to true.
## How to use `read_many_files` with Qwen Code
`read_many_files` searches for files matching the provided `paths` and `include` patterns, while respecting `exclude` patterns and default excludes (if enabled).
- For text files: it reads the content of each matched file (attempting to skip binary files not explicitly requested as image/PDF) and concatenates it into a single string, with a separator `--- {filePath} ---` between the content of each file. Uses UTF-8 encoding by default.
- The tool inserts a `--- End of content ---` after the last file.
- For image and PDF files: if explicitly requested by name or extension (e.g., `paths: ["logo.png"]` or `include: ["*.pdf"]`), the tool reads the file and returns its content as a base64 encoded string.
- The tool attempts to detect and skip other binary files (those not matching common image/PDF types or not explicitly requested) by checking for null bytes in their initial content.
Usage:
```
read_many_files(paths=["Your files or paths here."], include=["Additional files to include."], exclude=["Files to exclude."], recursive=False, useDefaultExcludes=false, respect_git_ignore=true)
```
## `read_many_files` examples
Read all TypeScript files in the `src` directory:
```
read_many_files(paths=["src/**/*.ts"])
```
Read the main README, all Markdown files in the `docs` directory, and a specific logo image, excluding a specific file:
```
read_many_files(paths=["README.md", "docs/**/*.md", "assets/logo.png"], exclude=["docs/OLD_README.md"])
```
Read all JavaScript files but explicitly include test files and all JPEGs in an `images` folder:
```
read_many_files(paths=["**/*.js"], include=["**/*.test.js", "images/**/*.jpg"], useDefaultExcludes=False)
```
## Important notes
- **Binary file handling:**
- **Image/PDF/Audio/Video files:** The tool can read common image types (PNG, JPEG, etc.), PDF, audio (mp3, wav), and video (mp4, mov) files, returning them as base64 encoded data. These files _must_ be explicitly targeted by the `paths` or `include` patterns (e.g., by specifying the exact filename like `video.mp4` or a pattern like `*.mov`).
- **Other binary files:** The tool attempts to detect and skip other types of binary files by examining their initial content for null bytes. The tool excludes these files from its output.
- **Performance:** Reading a very large number of files or very large individual files can be resource-intensive.
- **Path specificity:** Ensure paths and glob patterns are correctly specified relative to the tool's target directory. For image/PDF files, ensure the patterns are specific enough to include them.
- **Default excludes:** Be aware of the default exclusion patterns (like `node_modules`, `.git`) and use `useDefaultExcludes=False` if you need to override them, but do so cautiously.

259
docs/developers/tools/shell.md Normal file
View File

@@ -0,0 +1,259 @@
# Shell Tool (`run_shell_command`)
This document describes the `run_shell_command` tool for Qwen Code.
## Description
Use `run_shell_command` to interact with the underlying system, run scripts, or perform command-line operations. `run_shell_command` executes a given shell command, including interactive commands that require user input (e.g., `vim`, `git rebase -i`) if the `tools.shell.enableInteractiveShell` setting is set to `true`.
On Windows, commands are executed with `cmd.exe /c`. On other platforms, they are executed with `bash -c`.
### Arguments
`run_shell_command` takes the following arguments:
- `command` (string, required): The exact shell command to execute.
- `description` (string, optional): A brief description of the command's purpose, which will be shown to the user.
- `directory` (string, optional): The directory (relative to the project root) in which to execute the command. If not provided, the command runs in the project root.
- `is_background` (boolean, required): Whether to run the command in background. This parameter is required to ensure explicit decision-making about command execution mode. Set to true for long-running processes like development servers, watchers, or daemons that should continue running without blocking further commands. Set to false for one-time commands that should complete before proceeding.
## How to use `run_shell_command` with Qwen Code
When using `run_shell_command`, the command is executed as a subprocess. You can control whether commands run in background or foreground using the `is_background` parameter, or by explicitly adding `&` to commands. The tool returns detailed information about the execution, including:
### Required Background Parameter
The `is_background` parameter is **required** for all command executions. This design ensures that the LLM (and users) must explicitly decide whether each command should run in the background or foreground, promoting intentional and predictable command execution behavior. By making this parameter mandatory, we avoid unintended fallback to foreground execution, which could block subsequent operations when dealing with long-running processes.
### Background vs Foreground Execution
The tool intelligently handles background and foreground execution based on your explicit choice:
**Use background execution (`is_background: true`) for:**
- Long-running development servers: `npm run start`, `npm run dev`, `yarn dev`
- Build watchers: `npm run watch`, `webpack --watch`
- Database servers: `mongod`, `mysql`, `redis-server`
- Web servers: `python -m http.server`, `php -S localhost:8000`
- Any command expected to run indefinitely until manually stopped
**Use foreground execution (`is_background: false`) for:**
- One-time commands: `ls`, `cat`, `grep`
- Build commands: `npm run build`, `make`
- Installation commands: `npm install`, `pip install`
- Git operations: `git commit`, `git push`
- Test runs: `npm test`, `pytest`
### Execution Information
The tool returns detailed information about the execution, including:
- `Command`: The command that was executed.
- `Directory`: The directory where the command was run.
- `Stdout`: Output from the standard output stream.
- `Stderr`: Output from the standard error stream.
- `Error`: Any error message reported by the subprocess.
- `Exit Code`: The exit code of the command.
- `Signal`: The signal number if the command was terminated by a signal.
- `Background PIDs`: A list of PIDs for any background processes started.
Usage:
```bash
run_shell_command(command="Your commands.", description="Your description of the command.", directory="Your execution directory.", is_background=false)
```
**Note:** The `is_background` parameter is required and must be explicitly specified for every command execution.
## `run_shell_command` examples
List files in the current directory:
```bash
run_shell_command(command="ls -la", is_background=false)
```
Run a script in a specific directory:
```bash
run_shell_command(command="./my_script.sh", directory="scripts", description="Run my custom script", is_background=false)
```
Start a background development server (recommended approach):
```bash
run_shell_command(command="npm run dev", description="Start development server in background", is_background=true)
```
Start a background server (alternative with explicit &):
```bash
run_shell_command(command="npm run dev &", description="Start development server in background", is_background=false)
```
Run a build command in foreground:
```bash
run_shell_command(command="npm run build", description="Build the project", is_background=false)
```
Start multiple background services:
```bash
run_shell_command(command="docker-compose up", description="Start all services", is_background=true)
```
## Configuration
You can configure the behavior of the `run_shell_command` tool by modifying your `settings.json` file or by using the `/settings` command in the Qwen Code.
### Enabling Interactive Commands
To enable interactive commands, you need to set the `tools.shell.enableInteractiveShell` setting to `true`. This will use `node-pty` for shell command execution, which allows for interactive sessions. If `node-pty` is not available, it will fall back to the `child_process` implementation, which does not support interactive commands.
**Example `settings.json`:**
```json
{
"tools": {
"shell": {
"enableInteractiveShell": true
}
}
}
```
### Showing Color in Output
To show color in the shell output, you need to set the `tools.shell.showColor` setting to `true`. **Note: This setting only applies when `tools.shell.enableInteractiveShell` is enabled.**
**Example `settings.json`:**
```json
{
"tools": {
"shell": {
"showColor": true
}
}
}
```
### Setting the Pager
You can set a custom pager for the shell output by setting the `tools.shell.pager` setting. The default pager is `cat`. **Note: This setting only applies when `tools.shell.enableInteractiveShell` is enabled.**
**Example `settings.json`:**
```json
{
"tools": {
"shell": {
"pager": "less"
}
}
}
```
## Interactive Commands
The `run_shell_command` tool now supports interactive commands by integrating a pseudo-terminal (pty). This allows you to run commands that require real-time user input, such as text editors (`vim`, `nano`), terminal-based UIs (`htop`), and interactive version control operations (`git rebase -i`).
When an interactive command is running, you can send input to it from the Qwen Code. To focus on the interactive shell, press `ctrl+f`. The terminal output, including complex TUIs, will be rendered correctly.
## Important notes
- **Security:** Be cautious when executing commands, especially those constructed from user input, to prevent security vulnerabilities.
- **Error handling:** Check the `Stderr`, `Error`, and `Exit Code` fields to determine if a command executed successfully.
- **Background processes:** When `is_background=true` or when a command contains `&`, the tool will return immediately and the process will continue to run in the background. The `Background PIDs` field will contain the process ID of the background process.
- **Background execution choices:** The `is_background` parameter is required and provides explicit control over execution mode. You can also add `&` to the command for manual background execution, but the `is_background` parameter must still be specified. The parameter provides clearer intent and automatically handles the background execution setup.
- **Command descriptions:** When using `is_background=true`, the command description will include a `[background]` indicator to clearly show the execution mode.
## Environment Variables
When `run_shell_command` executes a command, it sets the `QWEN_CODE=1` environment variable in the subprocess's environment. This allows scripts or tools to detect if they are being run from within the CLI.
## Command Restrictions
You can restrict the commands that can be executed by the `run_shell_command` tool by using the `tools.core` and `tools.exclude` settings in your configuration file.
- `tools.core`: To restrict `run_shell_command` to a specific set of commands, add entries to the `core` list under the `tools` category in the format `run_shell_command(<command>)`. For example, `"tools": {"core": ["run_shell_command(git)"]}` will only allow `git` commands. Including the generic `run_shell_command` acts as a wildcard, allowing any command not explicitly blocked.
- `tools.exclude`: To block specific commands, add entries to the `exclude` list under the `tools` category in the format `run_shell_command(<command>)`. For example, `"tools": {"exclude": ["run_shell_command(rm)"]}` will block `rm` commands.
The validation logic is designed to be secure and flexible:
1. **Command Chaining Disabled**: The tool automatically splits commands chained with `&&`, `||`, or `;` and validates each part separately. If any part of the chain is disallowed, the entire command is blocked.
2. **Prefix Matching**: The tool uses prefix matching. For example, if you allow `git`, you can run `git status` or `git log`.
3. **Blocklist Precedence**: The `tools.exclude` list is always checked first. If a command matches a blocked prefix, it will be denied, even if it also matches an allowed prefix in `tools.core`.
### Command Restriction Examples
**Allow only specific command prefixes**
To allow only `git` and `npm` commands, and block all others:
```json
{
"tools": {
"core": ["run_shell_command(git)", "run_shell_command(npm)"]
}
}
```
- `git status`: Allowed
- `npm install`: Allowed
- `ls -l`: Blocked
**Block specific command prefixes**
To block `rm` and allow all other commands:
```json
{
"tools": {
"core": ["run_shell_command"],
"exclude": ["run_shell_command(rm)"]
}
}
```
- `rm -rf /`: Blocked
- `git status`: Allowed
- `npm install`: Allowed
**Blocklist takes precedence**
If a command prefix is in both `tools.core` and `tools.exclude`, it will be blocked.
```json
{
"tools": {
"core": ["run_shell_command(git)"],
"exclude": ["run_shell_command(git push)"]
}
}
```
- `git push origin main`: Blocked
- `git status`: Allowed
**Block all shell commands**
To block all shell commands, add the `run_shell_command` wildcard to `tools.exclude`:
```json
{
"tools": {
"exclude": ["run_shell_command"]
}
}
```
- `ls -l`: Blocked
- `any other command`: Blocked
## Security Note for `excludeTools`
Command-specific restrictions in `excludeTools` for `run_shell_command` are based on simple string matching and can be easily bypassed. This feature is **not a security mechanism** and should not be relied upon to safely execute untrusted code. It is recommended to use `coreTools` to explicitly select commands
that can be executed.

145
docs/developers/tools/task.md Normal file
View File

@@ -0,0 +1,145 @@
# Task Tool (`task`)
This document describes the `task` tool for Qwen Code.
## Description
Use `task` to launch a specialized subagent to handle complex, multi-step tasks autonomously. The Task tool delegates work to specialized agents that can work independently with access to their own set of tools, allowing for parallel task execution and specialized expertise.
### Arguments
`task` takes the following arguments:
- `description` (string, required): A short (3-5 word) description of the task for user visibility and tracking purposes.
- `prompt` (string, required): The detailed task prompt for the subagent to execute. Should contain comprehensive instructions for autonomous execution.
- `subagent_type` (string, required): The type of specialized agent to use for this task. Must match one of the available configured subagents.
## How to use `task` with Qwen Code
The Task tool dynamically loads available subagents from your configuration and delegates tasks to them. Each subagent runs independently and can use its own set of tools, allowing for specialized expertise and parallel execution.
When you use the Task tool, the subagent will:
1. Receive the task prompt with full autonomy
2. Execute the task using its available tools
3. Return a final result message
4. Terminate (subagents are stateless and single-use)
Usage:
```
task(description="Brief task description", prompt="Detailed task instructions for the subagent", subagent_type="agent_name")
```
## Available Subagents
The available subagents depend on your configuration. Common subagent types might include:
- **general-purpose**: For complex multi-step tasks requiring various tools
- **code-reviewer**: For reviewing and analyzing code quality
- **test-runner**: For running tests and analyzing results
- **documentation-writer**: For creating and updating documentation
You can view available subagents by using the `/agents` command in Qwen Code.
## Task Tool Features
### Real-time Progress Updates
The Task tool provides live updates showing:
- Subagent execution status
- Individual tool calls being made by the subagent
- Tool call results and any errors
- Overall task progress and completion status
### Parallel Execution
You can launch multiple subagents concurrently by calling the Task tool multiple times in a single message, allowing for parallel task execution and improved efficiency.
### Specialized Expertise
Each subagent can be configured with:
- Specific tool access permissions
- Specialized system prompts and instructions
- Custom model configurations
- Domain-specific knowledge and capabilities
## `task` examples
### Delegating to a general-purpose agent
```
task(
description="Code refactoring",
prompt="Please refactor the authentication module in src/auth/ to use modern async/await patterns instead of callbacks. Ensure all tests still pass and update any related documentation.",
subagent_type="general-purpose"
)
```
### Running parallel tasks
```
# Launch code review and test execution in parallel
task(
description="Code review",
prompt="Review the recent changes in the user management module for code quality, security issues, and best practices compliance.",
subagent_type="code-reviewer"
)
task(
description="Run tests",
prompt="Execute the full test suite and analyze any failures. Provide a summary of test coverage and recommendations for improvement.",
subagent_type="test-runner"
)
```
### Documentation generation
```
task(
description="Update docs",
prompt="Generate comprehensive API documentation for the newly implemented REST endpoints in the orders module. Include request/response examples and error codes.",
subagent_type="documentation-writer"
)
```
## When to Use the Task Tool
Use the Task tool when:
1. **Complex multi-step tasks** - Tasks requiring multiple operations that can be handled autonomously
2. **Specialized expertise** - Tasks that benefit from domain-specific knowledge or tools
3. **Parallel execution** - When you have multiple independent tasks that can run simultaneously
4. **Delegation needs** - When you want to hand off a complete task rather than micromanaging steps
5. **Resource-intensive operations** - Tasks that might take significant time or computational resources
## When NOT to Use the Task Tool
Don't use the Task tool for:
- **Simple, single-step operations** - Use direct tools like Read, Edit, etc.
- **Interactive tasks** - Tasks requiring back-and-forth communication
- **Specific file reads** - Use Read tool directly for better performance
- **Simple searches** - Use Grep or Glob tools directly
## Important Notes
- **Stateless execution**: Each subagent invocation is independent with no memory of previous executions
- **Single communication**: Subagents provide one final result message - no ongoing communication
- **Comprehensive prompts**: Your prompt should contain all necessary context and instructions for autonomous execution
- **Tool access**: Subagents only have access to tools configured in their specific configuration
- **Parallel capability**: Multiple subagents can run simultaneously for improved efficiency
- **Configuration dependent**: Available subagent types depend on your system configuration
## Configuration
Subagents are configured through Qwen Code's agent configuration system. Use the `/agents` command to:
- View available subagents
- Create new subagent configurations
- Modify existing subagent settings
- Set tool permissions and capabilities
For more information on configuring subagents, refer to the subagents documentation.

63
docs/developers/tools/todo-write.md Normal file
View File

@@ -0,0 +1,63 @@
# Todo Write Tool (`todo_write`)
This document describes the `todo_write` tool for Qwen Code.
## Description
Use `todo_write` to create and manage a structured task list for your current coding session. This tool helps the AI assistant track progress and organize complex tasks, providing you with visibility into what work is being performed.
### Arguments
`todo_write` takes one argument:
- `todos` (array, required): An array of todo items, where each item contains:
- `content` (string, required): The description of the task.
- `status` (string, required): The current status (`pending`, `in_progress`, or `completed`).
- `activeForm` (string, required): The present continuous form describing what is being done (e.g., "Running tests", "Building the project").
## How to use `todo_write` with Qwen Code
The AI assistant will automatically use this tool when working on complex, multi-step tasks. You don't need to explicitly request it, but you can ask the assistant to create a todo list if you want to see the planned approach for your request.
The tool stores todo lists in your home directory (`~/.qwen/todos/`) with session-specific files, so each coding session maintains its own task list.
## When the AI uses this tool
The assistant uses `todo_write` for:
- Complex tasks requiring multiple steps
- Feature implementations with several components
- Refactoring operations across multiple files
- Any work involving 3 or more distinct actions
The assistant will not use this tool for simple, single-step tasks or purely informational requests.
### `todo_write` examples
Creating a feature implementation plan:
```
todo_write(todos=[
{
"content": "Create user preferences model",
"status": "pending",
"activeForm": "Creating user preferences model"
},
{
"content": "Add API endpoints for preferences",
"status": "pending",
"activeForm": "Adding API endpoints for preferences"
},
{
"content": "Implement frontend components",
"status": "pending",
"activeForm": "Implementing frontend components"
}
])
```
## Important notes
- **Automatic usage:** The AI assistant manages todo lists automatically during complex tasks.
- **Progress visibility:** You'll see todo lists updated in real-time as work progresses.
- **Session isolation:** Each coding session has its own todo list that doesn't interfere with others.

54
docs/developers/tools/web-fetch.md Normal file
View File

@@ -0,0 +1,54 @@
# Web Fetch Tool (`web_fetch`)
This document describes the `web_fetch` tool for Qwen Code.
## Description
Use `web_fetch` to fetch content from a specified URL and process it using an AI model. The tool takes a URL and a prompt as input, fetches the URL content, converts HTML to markdown, and processes the content with the prompt using a small, fast model.
### Arguments
`web_fetch` takes two arguments:
- `url` (string, required): The URL to fetch content from. Must be a fully-formed valid URL starting with `http://` or `https://`.
- `prompt` (string, required): The prompt describing what information you want to extract from the page content.
## How to use `web_fetch` with Qwen Code
To use `web_fetch` with Qwen Code, provide a URL and a prompt describing what you want to extract from that URL. The tool will ask for confirmation before fetching the URL. Once confirmed, the tool will fetch the content directly and process it using an AI model.
The tool automatically converts HTML to text, handles GitHub blob URLs (converting them to raw URLs), and upgrades HTTP URLs to HTTPS for security.
Usage:
```
web_fetch(url="https://example.com", prompt="Summarize the main points of this article")
```
## `web_fetch` examples
Summarize a single article:
```
web_fetch(url="https://example.com/news/latest", prompt="Can you summarize the main points of this article?")
```
Extract specific information:
```
web_fetch(url="https://arxiv.org/abs/2401.0001", prompt="What are the key findings and methodology described in this paper?")
```
Analyze GitHub documentation:
```
web_fetch(url="https://github.com/QwenLM/Qwen/blob/main/README.md", prompt="What are the installation steps and main features?")
```
## Important notes
- **Single URL processing:** `web_fetch` processes one URL at a time. To analyze multiple URLs, make separate calls to the tool.
- **URL format:** The tool automatically upgrades HTTP URLs to HTTPS and converts GitHub blob URLs to raw format for better content access.
- **Content processing:** The tool fetches content directly and processes it using an AI model, converting HTML to readable text format.
- **Output quality:** The quality of the output will depend on the clarity of the instructions in the prompt.
- **MCP tools:** If an MCP-provided web fetch tool is available (starting with "mcp\_\_"), prefer using that tool as it may have fewer restrictions.

186
docs/developers/tools/web-search.md Normal file
View File

@@ -0,0 +1,186 @@
# Web Search Tool (`web_search`)
This document describes the `web_search` tool for performing web searches using multiple providers.
## Description
Use `web_search` to perform a web search and get information from the internet. The tool supports multiple search providers and returns a concise answer with source citations when available.
### Supported Providers
1. **DashScope** (Official, Free) - Automatically available for Qwen OAuth users (200 requests/minute, 2000 requests/day)
2. **Tavily** - High-quality search API with built-in answer generation
3. **Google Custom Search** - Google's Custom Search JSON API
### Arguments
`web_search` takes two arguments:
- `query` (string, required): The search query
- `provider` (string, optional): Specific provider to use ("dashscope", "tavily", "google")
- If not specified, uses the default provider from configuration
## Configuration
### Method 1: Settings File (Recommended)
Add to your `settings.json`:
```json
{
"webSearch": {
"provider": [
{ "type": "dashscope" },
{ "type": "tavily", "apiKey": "tvly-xxxxx" },
{
"type": "google",
"apiKey": "your-google-api-key",
"searchEngineId": "your-search-engine-id"
}
],
"default": "dashscope"
}
}
```
**Notes:**
- DashScope doesn't require an API key (official, free service)
- **Qwen OAuth users:** DashScope is automatically added to your provider list, even if not explicitly configured
- Configure additional providers (Tavily, Google) if you want to use them alongside DashScope
- Set `default` to specify which provider to use by default (if not set, priority order: Tavily > Google > DashScope)
### Method 2: Environment Variables
Set environment variables in your shell or `.env` file:
```bash
# Tavily
export TAVILY_API_KEY="tvly-xxxxx"
# Google
export GOOGLE_API_KEY="your-api-key"
export GOOGLE_SEARCH_ENGINE_ID="your-engine-id"
```
### Method 3: Command Line Arguments
Pass API keys when running Qwen Code:
```bash
# Tavily
qwen --tavily-api-key tvly-xxxxx
# Google
qwen --google-api-key your-key --google-search-engine-id your-id
# Specify default provider
qwen --web-search-default tavily
```
### Backward Compatibility (Deprecated)
⚠️ **DEPRECATED:** The legacy `tavilyApiKey` configuration is still supported for backward compatibility but is deprecated:
```json
{
"advanced": {
"tavilyApiKey": "tvly-xxxxx" // ⚠️ Deprecated
}
}
```
**Important:** This configuration is deprecated and will be removed in a future version. Please migrate to the new `webSearch` configuration format shown above. The old configuration will automatically configure Tavily as a provider, but we strongly recommend updating your configuration.
## Disabling Web Search
If you want to disable the web search functionality, you can exclude the `web_search` tool in your `settings.json`:
```json
{
"tools": {
"exclude": ["web_search"]
}
}
```
**Note:** This setting requires a restart of Qwen Code to take effect. Once disabled, the `web_search` tool will not be available to the model, even if web search providers are configured.
## Usage Examples
### Basic search (using default provider)
```
web_search(query="latest advancements in AI")
```
### Search with specific provider
```
web_search(query="latest advancements in AI", provider="tavily")
```
### Real-world examples
```
web_search(query="weather in San Francisco today")
web_search(query="latest Node.js LTS version", provider="google")
web_search(query="best practices for React 19", provider="dashscope")
```
## Provider Details
### DashScope (Official)
- **Cost:** Free
- **Authentication:** Automatically available when using Qwen OAuth authentication
- **Configuration:** No API key required, automatically added to provider list for Qwen OAuth users
- **Quota:** 200 requests/minute, 2000 requests/day
- **Best for:** General queries, always available as fallback for Qwen OAuth users
- **Auto-registration:** If you're using Qwen OAuth, DashScope is automatically added to your provider list even if you don't configure it explicitly
### Tavily
- **Cost:** Requires API key (paid service with free tier)
- **Sign up:** https://tavily.com
- **Features:** High-quality results with AI-generated answers
- **Best for:** Research, comprehensive answers with citations
### Google Custom Search
- **Cost:** Free tier available (100 queries/day)
- **Setup:**
1. Enable Custom Search API in Google Cloud Console
2. Create a Custom Search Engine at https://programmablesearchengine.google.com
- **Features:** Google's search quality
- **Best for:** Specific, factual queries
## Important Notes
- **Response format:** Returns a concise answer with numbered source citations
- **Citations:** Source links are appended as a numbered list: [1], [2], etc.
- **Multiple providers:** If one provider fails, manually specify another using the `provider` parameter
- **DashScope availability:** Automatically available for Qwen OAuth users, no configuration needed
- **Default provider selection:** The system automatically selects a default provider based on availability:
1. Your explicit `default` configuration (highest priority)
2. CLI argument `--web-search-default`
3. First available provider by priority: Tavily > Google > DashScope
## Troubleshooting
**Tool not available?**
- **For Qwen OAuth users:** The tool is automatically registered with DashScope provider, no configuration needed
- **For other authentication types:** Ensure at least one provider (Tavily or Google) is configured
- For Tavily/Google: Verify your API keys are correct
**Provider-specific errors?**
- Use the `provider` parameter to try a different search provider
- Check your API quotas and rate limits
- Verify API keys are properly set in configuration
**Need help?**
- Check your configuration: Run `qwen` and use the settings dialog
- View your current settings in `~/.qwen-code/settings.json` (macOS/Linux) or `%USERPROFILE%\.qwen-code\settings.json` (Windows)