canon-keeper-mcp

# Canon Keeper MCP Server An MCP (Model Context Protocol) server that persists learnings from Copilot conversations to `copilot-instructions.md`. **No API key required!** Copilot extracts learnings using its own LLM, this tool handles deduplication and formatting. ## Quick Install ```bash pip install canon-keeper-mcp python -m canon_keeper_mcp install # Or specify a workspace python -m canon_keeper_mcp install --workspace /path/to/project ``` The installer will: 1. ✅ Configure `.vscode/mcp.json` with the canon-keeper server 2. ✅ Enable MCP in `.vscode/settings.json` 3. ✅ Create/update `.github/copilot-instructions.md` with: - Memory Persistence Protocol (`@History` directive) - Best practices template - Session Learnings Log table ## Features - **Zero Configuration**: No API keys needed - uses Copilot's built-in LLM - **Smart Deduplication**: Uses Jaccard similarity to avoid duplicate entries - **Simple Integration**: Works with Copilot's existing capabilities - **Markdown Formatting**: Returns properly formatted table rows for the Session Learnings Log ## Usage After installation, reload VS Code and use these trigger phrases in any Copilot chat: ``` User: @History save what we learned User: save this learning User: remember this User: add to memory ``` Copilot will: 1. Extract learnings from the conversation (using its own LLM) 2. Call the MCP tool to deduplicate against existing entries 3. Append new learnings to your `copilot-instructions.md` 4. Report what was saved/skipped ## Manual Installation If you prefer to install manually: ### 1. Install the package ```bash pip install canon-keeper-mcp ``` ### 2. Configure MCP server Add to `.vscode/mcp.json`: ```json { "servers": { "canon-keeper": { "type": "stdio", "command": "/path/to/your/python", "args": ["-m", "canon_keeper_mcp"] } } } ``` > **Note:** Replace `/path/to/your/python` with the absolute path to your Python executable (e.g., `C:/Users/You/p…