Codesteward Review
Agentic code review that knows your graph.
Gate every merge. Steward every branch. Self-hosted — your cloud, your models, your keys.
Codesteward Review is distributed under Apache-2.0 for self-host installs. There is no hosted SaaS product today — this documentation is for teams evaluating and operating their own deployment.
Most AI review tools skim a diff and guess. Codesteward runs multi-agent reviews against a structural code graph (call chains, dependencies, auth paths) so findings are grounded in how the codebase actually works — not only the patch text.
For engineering leaders
Consistent merge gates, durable findings, policy from the base branch, and an audit trail you can show compliance.
For platform / DevOps
Compose demo in minutes, Helm for production, Keycloak OIDC, horizontal workers, optional KEDA queue scaling.
For developers
PR comments and check runs that cite structure, not noise — plus a product UI for sessions, findings, and learnings.
Two review modes, one platform
| Gate | Stewardship | |
|---|---|---|
| When | PR / MR open, push, or @codesteward review | Long-lived branches and path scopes |
| Scope | Diff-focused units | Package / path / tree batches |
| Output | Inline review, check run, gate verdict | Durable findings lifecycle |
| Policy | STEWARD.md + path rules from base branch | Same model |
What you get
- Graph-aware specialists — correctness, security, testing, rules, and more, with optional thorough discourse
- Product UI — sessions, findings, reports, connectors, models, org settings, platform ops
- CLI + GitHub Action — same pipeline outside the browser
- Learning loop — 👍/👎 and dismissals become quieter next reviews
- Multi-SCM — GitHub App, GitLab, Bitbucket, Azure DevOps, Gitea
- Multi-org ready — Keycloak groups, RBAC, optional SCIM; worker isolation for shared hosts
Decide in three steps
- Why Codesteward — when graph-aware review is worth running yourself
- Quick start — category Compose stack on your laptop
- Kubernetes quick start or Install overview — Helm / Compose, Postgres, Keycloak, workers
Then explore the UI guide and review pipeline when you want depth.
License & cost model
- Software: Apache-2.0 (see repository
LICENSE) - Your cost: infrastructure + LLM API usage (OpenAI, Anthropic, SpaceXAI, OpenAI-compatible, LiteLLM, …)
- Not included: managed multi-tenant SaaS, hosted billing portal for end customers
Ready to try? → Quick start