Knos MCP server

Knos is a memory and coordination helper for people who run coding agents on their computer. It remembers what was decided in past sessions and keeps two agents from stepping on each other's work. It is handy if you use more than one AI coding tool, or work with other people on the same code.

What this MCP server does

You ask your AI a question, or you let it start editing a file. The AI uses this helper to talk to Knos on your machine. Knos looks through past sessions, commits, and project notes, or checks whether another agent already claimed that file. Then it answers you, or it blocks the edit and tells you who is working there. Everything happens locally, so nothing leaves your computer unless you set up a team.

Click to zoom

What you can do with it

• Ask why a past decision was made and get an answer with sources
• Claim a file or folder so other agents cannot edit it
• See which agents and machines are protected right now
• Set a daily spending limit for an agent
• Check what an agent claimed, finished, or abandoned
• Keep secrets like .env and .ssh files away from your agents

Try asking your AI

• “Why did we drop redis?”
• “Claim the parser folder so nobody else touches it”
• “What did the codex agent work on yesterday?”
• “Run knos doctor and tell me which agents are guarded”

What it gives back to you

You get plain answers in the chat, with the source of each one, like a past session or a commit. For claims, you get a short confirmation or a refusal that names who holds the file. For status checks, you get a list of agents and machines and whether they are guarded. Nothing is written to a website; it all stays on your machine.

Before you start

Who it's for

Developers and small teams who run several AI coding agents and want them to share memory and avoid collisions.