Build in-house style.
Agents inherit the design and architecture patterns already in your code, so new work matches what's there — not a generic default.
The navigation star for coding agents
Rigel charts every service, API and event across your repos — tens to hundreds of them — briefs any coding agent over MCP, and scores every pull request, so a change in one repo doesn't break another.
Works today · GitHub · Node.js / TypeScript · Claude on your key or AWS · self-host Beta
Built for microservice estates
Microservice architectures span tens to hundreds of repositories — no single coding session can hold that whole system in context, and each repo's agent sees only its own repo. An agent renames a field in a shared package; the reviewer sees one repo, not the two services in other repos that read it. Rigel holds the whole chart and briefs the agent first.
orders-api.
Nobody sees who consumes @acme/orders-client. The rename ships.
find_consumers over MCP.
Two services in two other repos depend on the package. The agent keeps customerId @deprecated and names both in the PR.
Matching is package-level today — Rigel reads package.json manifests. Endpoint & queue matching Beta · consumers in the Impact PR comment Coming soon
One source of truth
Rigel centralizes what a distributed system knows about itself — service topology, API and event contracts, and the patterns your teams actually follow — and serves it to any coding agent over MCP. An agent working on one service builds consistently with the other hundred, instead of guessing from a single repo.
Agents inherit the design and architecture patterns already in your code, so new work matches what's there — not a generic default.
APIs, events and queues across repos are mapped, so a change honors the contracts other services and workflows depend on.
How your services are configured, built and deployed travels with the chart, so agents follow your conventions by default.
How it works
New here? The 30-second version.
Rigel reads all your repos and draws one branch-aware map of how your services connect — refreshed on every merge.
Before an AI coding agent changes code, it asks Rigel what else depends on it, over MCP.
In repos where you turn Impact on, every pull request gets an explainable risk score and blast radius.
The PRs Rigel's own agents open are tested, reviewed and checked against your rules, then merged inside your branch protection — or held for a person, with a reason.
The platform
// built from your code · branch-aware · refreshed on every merge
A live, branch-aware map of services, APIs and events, built from code and queryable by any agent over MCP.
Every pull request scored for cross-service risk before it merges — the consumers a change will break, named.
The PRs Rigel's own agents open merge only when your CI, review and policy gates pass. Everything else waits.
Fit & proof
Rigel is in Beta with design partners: engineering orgs running 20+ services across many repos. These are mechanisms you can check, not benchmarks we're asking you to take on faith.
Design partners get early access, a direct line to the team that builds Rigel, and a say in the roadmap. We'll publish results here as partners agree to share them.
Go deeper
Architecture, the risk formula, and what ships today.
Security & trust →Defaults, data flows and honest disclosures.
Connect over MCP →20 read-only tools any MCP client can call.
Enterprise & self-host →Run Rigel in your own cloud, and rollout.
Pricing →How plans and the design-partner program work.
Why Rigel →The case for cross-repo context and gates.
Engineers
Trace one of your repos from onboarding to a governed PR.
Design partners: engineering orgs with 20+ services across many repos, on GitHub, with Node.js / TypeScript services today.