The idea
The work can be distributed. Your attention should not be.
Claude Code in one terminal. Codex CLI on the tests. Cursor on the review. Captain Code already runs those AI agents. Gouvernail is the desk that keeps every project captain in view, and asks you only when a person has to decide.
Captain Code runs the agents. Gouvernail keeps the desk.
Captain Code chooses the AI agent, runs the tests and keeps the task record. It works on its own. Gouvernail reaches those captains from the web or WhatsApp, and keeps the decisions your institution has to make.
A persistent project relationship
A captain knows the project’s purpose and backlog. You should be able to reach it from a terminal, the web or WhatsApp, without rebuilding context each time.
A useful question, not another alert
“What should I do?” is not enough. A captain should explain the recommendation, evidence, uncertainty and consequences. Routine progress belongs in the work view; questions belong in your decision inbox.
Captains that work together
Related projects depend on each other. Captains should ask each other for help inside the limits you set, and come to you only when a step is unclear or needs authority.
Responsibility after delivery
Work stays connected to its inputs after it is shared. When an input is withdrawn, the next question is which results rely on it and how to repair them. A reference alone is not the same as a dependency.
Each institution keeps authority
A shared website does not mean shared permissions. People act in an institutional role. Captains enforce execution limits. Contribution checks govern reuse. A delivered message cannot bypass those boundaries.
This is the direction we are building toward. It is not a claim of proven savings or complete security.
What needs to happen first ↗