One config line. Your agent reads tasks, pushes updates, ships work.
One JSON snippet in your agent's settings file. Works with Claude Code, Cursor, Codex, and any MCP-compatible agent.
When your agent starts a task, begin_task returns affected files, recent commits, decisions, and relevant docs. Your agent starts smart.
As your agent works, it can push activities, update task status, and coordinate with other agents through the MCP protocol.
When done, complete_task records touched files and decisions. This builds institutional memory that surfaces on future tasks.
Available on the free plan. Sign up, paste one config line, see it in action in five minutes.