IDE Integration¶

WeiLink integrates with AI coding assistants so you can receive and send WeChat messages directly from your coding session. The weilink setup command automates the installation of hooks, MCP server configuration, and slash commands.

pip install weilink[mcp]

Supported Assistants¶

Claude Code¶

weilink setup claude-code

This creates a symlink at ~/.claude/plugins/weilink/ pointing to the bundled integration assets. After installation, restart Claude Code to activate.

What gets installed¶

How auto-polling works¶

sequenceDiagram
    participant U as User
    participant CC as Claude Code
    participant H as poll_inject.py
    participant DB as messages.db

    U->>CC: Submit prompt
    CC->>H: UserPromptSubmit hook
    H->>DB: Query messages since last check
    DB-->>H: New messages (if any)
    H-->>CC: {"additionalContext": "== New WeChat Messages ==\n..."}
    CC->>U: Response (with message context)

Every time you submit a prompt, the UserPromptSubmit hook runs weilink hook-poll to check the local SQLite message store for new messages received since the last check. If there are new messages, they are injected into Claude's context as additionalContext. This is a local-only operation (no network call) and typically completes in under 100ms.

Options¶

Codex (OpenAI)¶

weilink setup codex

Installs a hook script, merges hook registration into ~/.codex/hooks.json, and copies the /weilink slash command.

What gets installed¶

MCP registration¶

The MCP server must be registered separately after setup:

codex mcp add weilink -- weilink mcp

Options¶

OpenCode¶

weilink setup opencode

Adds the WeiLink MCP server to ~/.config/opencode/opencode.json and installs the /weilink slash command.

What gets installed¶

Options¶

Manual Configuration¶

For other tools (Cursor, VS Code, Claude Desktop, etc.), configure the MCP server manually. See MCP Server — Client Configuration for examples.

To add auto-polling to a tool that supports UserPromptSubmit hooks, create a hook that runs:

weilink hook-poll

The command outputs JSON:

{"has_messages": true, "count": 3, "context": "== New WeChat Messages ==\n..."}

Wrap the context value as {"additionalContext": "<context>"} in your hook's stdout.

weilink hook-poll¶

Internal command used by hook scripts. Reads the local SQLite message store for messages received since the last check.

weilink hook-poll                # Poll for new messages
weilink hook-poll --limit 50     # Return up to 50 messages
weilink hook-poll --reset        # Clear poll state and exit

State is tracked in ~/.weilink/.hook_state.json. On first run (or after --reset), it looks back 5 minutes.