Cockpit by RiftSeed

In development. Cockpit is where the projects are developed: code agents work through a protocol (session briefing, requests checked against requirements, impact analysis, tests), and a person confirms the requirements and decides every release.

A prototype in development, used every day on the lab’s own projects. Early access on request.

What it does today

• A map for every project: The parts of the product coloured by the tests that protect them, the requirements that are cited and those that are not, the files no requirement explains.
• Requests and requirements: Every request keeps the words of whoever made it and is checked against the existing requirements, also with a decision model; a person confirms the requirement.
• Gate and ratchets: The CI result on every push and the ratchet ceilings, held by the server: a ceiling goes up only with a recorded decision.
• Verdict and incidents: A release waits for a person’s verdict, and open incidents block it. An incident closes only with the fix commit and the guard test.
• The studio and the Commander: Who holds what, what waits for a person. The Commander, an AI technical program manager, assigns tasks that agents work on a separate branch, with a draft pull request: never a deploy.
• The report: The project explained to non-technical readers: where it stands, what is ready, what is missing and who has to do what.

The journey of a request

The same steps as the diagram on /method, with what holds each of them in Cockpit.

1. Request: The request is logged in the words of whoever made it and checked against the requirements that already exist.
2. Requirement: Agents propose requirements; only a person confirms them, in Cockpit.
3. Impact: Before touching the code the agent asks for the impact analysis: files, modules and tests to run, each with its reason.
4. Agents: Every session starts from a briefing (priorities, open requirements, failing tests) and ends with commits, files and tests on record.
5. Gate: The full test run on every push; the ratchet ceilings are held by the server and can only go down.
6. Person: The verdict on the release comes from a person, in Cockpit or in the session: no agent approves.
7. Release: Every release stays on record with its version, its verdict and its notes.
8. Probes: A failure becomes an incident, which closes only with the guard test that reproduces it.

Which agents

Claude Code and Codex connect with a token, through MCP. A hook opens every session with the briefing and closes it by asking for the session log; while a task assigned by the Commander is held, a local guard denies deploys, merges and access to credentials.

What it does not do yet

• One person uses it: there is the administrator’s access, not yet organisations or teams.
• There is no installable package: connecting another repository means copying the hook kit by hand.
• Assurance levels computed from the evidence and structural code coverage do not exist yet; coverage today is per requirement.
• No price and no licence: it is a prototype.

Early access

For people who develop with code agents and want to try it on a project of their own. Write in the site chat: no price, no promised date. Contact

The method · RiftSeed and no-code platforms