The planned experience
From watching agents to steering projects.
Captain Code runs the AI agents. Gouvernail is where you see every project captain and answer the decisions. Institutions stay separate. A message is not permission to reuse someone else’s work.
- Agents
Captain Code runs the AI agents
Each project captain uses Captain Code to run Claude Code, Codex CLI, Cursor and the API models you connect. Gouvernail does not replace those agents. It shows their work and brings you the decisions.
- Fleet
One captain per project
See every captain in one place: where it runs, its current task, when it was last seen, and whether it needs you. A terminal is a view into a project, not its identity.
- Decisions
The recommendation, with the reason
Each update says what was done, what was learned, what is left and what is still uncertain. A decision comes with a recommendation, its reason and its consequences. Open the evidence before you decide.
- Coordination
Captains ask each other, you break ties
A captain can send a request to a linked project’s captain through Gouvernail. Each request carries its sender, a hop limit and a budget. Questions get answers; changes, releases and spending wait for a person.
- Backlog
Done, current, next and blocked
Keep a cursor on the project’s own tracker. See why an item is next and which dependency stops it.
- Access
Web, WhatsApp and terminal
Work in the terminal. See the fleet on the web. Answer low-risk questions on WhatsApp. High-risk approvals need a stronger channel.
- Cloud captains
Start one from the web
Start a captain in the cloud when a project needs one. Cloud captains use your organization’s API accounts. Personal subscription logins stay on your own computers.
- Execution
Sandboxed remote work
Remote code and tests run in a prepared OpenShell sandbox. With no sandbox ready, the task waits or is refused. There is no silent fallback to your computer.
- Collaboration
Repair what actually depends on a change
Separate messages from dependencies and citations. A withdrawn input pauses the work that builds on it; a citation gets a notice and unrelated work continues.
Who does what
Gouvernail connects people to their captains and keeps the decisions. Captain Code runs the work: models, workers, tests and budgets. Captain Code is open source and works without Gouvernail. captaincode.ai ↗
What is already in the codebase
Project chat, work units, human decision triggers, and web and WhatsApp channels are the starting point.
The direct coding-agent runner is still present. It retires after captains can take over its work and pass the security, migration and recovery checks.
See the build sequence ↗