Trinity Lite MCP server

Trinity Lite is a small helper that keeps track of coding tasks you give to command line AI agents like Codex and Claude Code. It stores every task and result in one local file on your computer, so you can see what happened and recover if something disconnects. It is handy if you already run one or two of these agents in your terminal and want them to work together without copy-pasting between windows.

What this MCP server does

You ask your AI to send a coding task, for example to implement a rate limiter. Your AI uses this helper, and the helper picks the right agent based on tags like implement or review. The agent runs the task in your terminal and writes the result back into a local database. Your AI then reads that result and shows it to you in the chat. If a client disconnects, you can ask your AI to look up the latest task and its status.

Click to zoom

What you can do with it

• Send a coding task to a specific agent like Codex or Claude Code
• Let the router pick the right agent based on the task
• Check the status and result of any task by its ID
• Recover the latest task if your chat disconnected
• List recent tasks, filtered by agent
• Run one worker cycle for an agent on demand
• Read a task's full evidence chain as a proof bundle

Try asking your AI

• “Send a task to implement a rate limiter for the API”
• “Dispatch a parser task and let the router choose the agent”
• “What is the status of my last task?”
• “Show me the proof bundle for task 42”

What it gives back to you

You get back plain answers in the chat: task IDs, statuses like pending or accepted, and the result text the agent produced. You can also ask for a list of recent tasks or a proof bundle that shows the route, review, verification, and acceptance steps. If a stage was never recorded, it says so instead of making something up.

Before you start

Who it's for

Developers who already run two or more command line coding agents and want a simple, local way to hand off work and keep a record.