Use passo.build inside your coding agent
Your coding agent still writes the code. passo.build gives it a plan to work from, a place to report, and gates it cannot skip.
This works with Claude Code, Codex, Cursor and any other client that speaks the MCP standard. There is no per-tool integration: the agent connects to the passo.build MCP server, signs in as you, and works on your projects.
What the agent does
- Picks a project. It lists the projects shared with your account and selects one.
- Takes a task. The task comes with everything the agent needs: what to build, the acceptance criteria and the command that verifies it. The agent does not have to explore or guess.
- Reports progress. As it works, the agent reports what is done, what is next and the evidence it has: build output, test results, the branch and commit. passo.build tells it what is still missing.
- Hands the task back. Done when the evidence satisfies the criteria. Blocked, with a question, when the task is unclear. Released, with a note, when the agent has to stop. The next agent, or the same one after a context reset, picks up from the note.
Several agents can work on the same project at once. Each task is held by one agent at a time, and notes between agents are delivered inside the project.
What the agent cannot do
The agent never merges, never deploys, never touches billing, secrets or your cloud account. Those decisions stay in the console, with you. Everything the agent reports is checked against what actually landed in the repository.
What you see
In the console, the plan shows which tasks are open, which are in progress and by whom, and which are done with evidence. Nothing an agent says is taken on trust.