Vasilije Markovic

Add memory to your coding copilot

I just deployed the latest version of the cognee MCP (Model Context Protocol) server locally. It's a standalone thing that talks to copilots and can actually work with your local code, notes, system files, etc.\

A few new things we added:

SSE support is finally in. You can now use --transport sse for real-time updates instead of just stdio. Makes integration feel way smoother.

Built-in logging is solid. Logs go to a file by default, but you can see stuff in real-time if you're running it locally. Helpful for debugging weird pipeline issues.

It can read local files directly (like markdown, source files, even .cursor/rules). Async processing and status tracking are built in.

Experimental support for Cursor rules is also there (only in dev branch though). It automatically pulls in rules from .cursorrules or .cursor/rules—kind of cool for enforcing project-specific logic.

If you're tinkering with local AI assistants or want more control over what your copilot knows, it’s worth trying out.

https://docs.cognee.ai/how-to-guides/deployment/mcp

7 views

Add a comment

Replies

Be the first to comment