Review coverage that scales with your agents.
Agents ship faster than anyone can review, so things slip. A steward is a narrow owner for one dimension of your codebase — auth, billing, performance — that watches every change and catches what matters. Add stewards as you grow, from one repo to a whole team.
Other AI reviewers optimize precision, to protect your attention. Stewards assume your agents already fix what review finds, so they optimize the one thing left: missing nothing. You don’t install a steward, you build a fleet, a composable layer for your agent loops and software factories.
A steward is an owner, not another reviewer.
If it would benefit from an owner, it will benefit from a steward. Below: real stewards, running in real repos.
How a steward works
Before the agent builds
Ask the event reliability steward about a telemetry plan. It spots the part that needs a test before your agent writes the first diff.
After the diff lands
The same steward reads the changed files, raises the concern that still needs you, then clears the check when the agent proves the fix.
Install where your agents already work
Use a coding agent that supports remote HTTP MCP servers and browser sign-in. Connect it directly to Steward:
https://mcp.steward.foo- Add this URL in your agent’s MCP server settings.
- Complete the browser sign-in with the same Steward account you use on this site.
- Open the repository in your agent. In your Steward workspace, choose New steward, select the repository, describe the duties, and copy the prompt into your agent.
The agent reads the server’s current tool descriptions and schemas, proposes a focused steward, and previews it with you before saving.