Launching today

ctxnest
The Local Brain & Hands for Your AI Coding Agents
4 followers
The Local Brain & Hands for Your AI Coding Agents
4 followers
Your AI coding agent is a goldfish. Great at reasoning, zero memory, no safe way to act. CtxNest fixes that with a local Brain/Hands/Eyes loop over MCP β persistent knowledge vault, sandboxed command execution, and a feedback journal that compounds every session. Works with Claude Code, Cursor, Gemini, Codex. 100% local, 0% egress. Your data never leaves your SSD.







I kept hitting the same wall with every AI coding assistant I tried. They're brilliant in the moment. But close the tab and they forget everything β your architecture, your standards, what you decided last week and why. Every new session meant 10 minutes of re-explaining before the agent could actually help.
The other problem: I wanted my agents to do things, not just suggest things. But letting an AI run arbitrary commands felt like handing a stranger your keyboard. So I'd babysit every execution manually, which defeated the whole point.
And the context pollution β generated docs, migration plans, decision logs, all committed alongside real code with no separation. My repos were becoming half AI noise.
So I built the thing I needed.