Bring Rigel to your team.
- For your executives /solutions#executives
- For your VP Engineering /solutions#platform-leaders
- For your security reviewer /security
- For your product owners /solutions#product-owners
RigelSolutions
One chart. Every role that needs it.
Whoever you are in the room, Rigel answers the question you're asking about coding agents. Pick your seat.
Your seat
Executives
VP Engineering & Platform
Regulated industries
Staff engineers & architects
Product owners
Each role has its own link. Send it on.
Role · Executives
Use AI coding agents without breaking other teams' systems.
The pain — Your teams are adopting coding agents. Each agent sees one codebase. When a change breaks a service another team owns, customers notice first.
Risk reduction Rigel Impact →
Agents check what depends on a package before they change it, and PRs carry a published risk score.
Delivery speed Rigel Autopilot →PRs Rigel's own agents open can merge through your gates without waiting in a queue.
Cost control Pricing →Model usage runs on your own account, with a dollar cap on every run.
Adoption effort Rollout →Start with one application's repos, then expand team by team.
Control Security →Only changes that pass your rules merge on their own; the rest wait for a person, with a reason.
See it on a sample estate, or on your own repos.
Role · VP Engineering & Platform
Fewer cross-team breaks. Less senior review time.
The pain — Agents open PRs faster than you can review them, and nothing says which ones may merge themselves.
Rigel centralizes cross-repo application knowledge, workflows, architecture and design patterns and infra conventions, so agents across a microservice estate — tens to hundreds of repos — build consistently with the other hundred services instead of guessing from a single repo.
1
Fewer incidents from cross-service changes.
Agents ask Rigel who depends on a package before editing, and PRs carry a risk score.
2
Routine agent PRs off your reviewers' plate.
Rigel's patch, minor and docs PRs merge only when CI, review and gates pass; everything else is held with a reason.
3
Agent spend you can see and cap.
Every run records cost and has a USD cap.
Pilot one application's repos, then expand team by team. See what adoption takes →
Delivered by
Rigel Autopilot →
PR Supervisor · Repo Guard · Flight Log
Rigel Impact →Blast-radius risk score on PRs you choose
Rigel Map →Deep Survey · Constellations · Living Docs
How the merge policy works
Rigel's own agents (bug fixes, dependency upgrades, docs refreshes) open PRs that the supervisor gates by origin: semver and path gates for dependency bumps, a fixer loop capped at 4 iterations and a fresh review for code changes, and a [needs-review] brake in the title.
A Repo Guard you install in each repo blocks Rigel's bot PRs that touch the wrong paths. Humans are never blocked. PRs from your other coding agents get a blast-radius risk score when Impact is on for the repo.
GitHub today. Code-change auto-merge requires GitHub Copilot code review.
| Time | What happens |
|---|---|
| 08:40 | Dependabot flags a vulnerable package in orders-api. The Dependency Upgrade Advisor opens one tested PR. |
| 09:20 | Merged on course. Patch bump, lockfile only, CI green. semver ✓paths ✓CI ✓threads ✓ |
| 11:05 | Held off course. A second PR bumps an external major. The decision comment names the gate.semver: external majorwaiting for a human |
| 14:30 | Another team's coding agent opens a PR. Impact posts a risk score with every factor and the impact chain. |
| 17:00 | The Flight Log has one Slack thread per agent PR. Costs roll up in the admin view. |
Sample repo and times. Illustration.
Bring the policy to your platform team.
Role · Regulated industries
Agents near sensitive systems, under hard controls.
For CTOs in healthcare, fintech and insurance.
The pain — You can't ship sensitive data to a vendor's cloud, and you can't adopt agents you can't explain to an auditor.
1
Your cloud, your model account. Beta
Self-host Rigel Enterprise on your own Postgres. Model calls use your Anthropic key or Claude through your AWS account.
2
The agent comes to the data. Beta
A self-hosted agent host runs next to your database, behind a read-only boundary with default-deny column redaction. Data critics fail closed.
3
Every decision on the record.
Runs, supervisor decisions, admin cross-user reads, agent-to-agent hops and sensitive-database queries are logged.
Delivered by
Security & Trust →
Data flows · guardrails · operator duties
Rigel Enterprise →Self-host in your cloud Beta
Rigel Impact →Release Analysis · CAB pack from the real diff
How the controls work
Agents can't edit their own checks. Rigel's code-change sandbox refuses changes to .github/** three ways: at write time, on a pre-push diff and in a git pre-push hook. Pair it with a GitHub ruleset for defence in depth.
A CAB pack from the real diff. Release Analysis compares production with staging and writes the change ticket, checklist and data-engineering change notice, for release pipelines on AWS.
Embeddings disclosure. Code excerpts go to your chosen model provider and to Voyage AI for embeddings, under your own Voyage key. Coming soon A unified audit export.
No certifications claimed. Rigel holds no compliance certifications today. We describe the specific mechanisms so your security team can assess them directly. Full data-flow disclosure →
| Time | What happens |
|---|---|
| 10:10 | A product owner asks how many claims stall in review. The question goes to the agent host beside the claims database. Beta |
| 10:10 | It queries through the read-only boundary. Columns your policy doesn't allow come back redacted. READ ONLYrow cap member_name: [REDACTED] |
| 10:11 | Only the requirement-level answer leaves. A fail-closed critic checks it first. The hop log keeps intent and verdict, never the values. |
| 13:40 | A fixer agent decides a CI check is wrong. The write is refused, so it files a finding with a proposed diff for a human..github/workflows/ci.yml: refused |
| 16:00 | Release Analysis compares production with staging and drafts the CAB pack from the real diff. |
Sample data and times. Illustration.
Walk your security team through it.
Role · Staff engineers & architects
Stop being the human dependency graph.
The pain — Every contract change and every design review routes through the map in your head.
1
Your coding agent knows the consumers.
Rigel MCP gives any MCP client 20 read-only tools, including find_consumers, cross_repo_impact and seven code-graph tools.
2
Design reviews that don't need you.
Ask Rigel answers system questions with cited sources. Solution Course drafts designs grounded in the chart.
3
Architectural problems caught on the PR.
Blast Radius puts a transparent 8-factor risk score and impact chain on PRs. Persona review checks how the service is meant to work.
Delivered by
Rigel MCP →
20 read-only tools · stdio or HTTP
Rigel Map →Ask Rigel · Constellation Assembly · code graph
Rigel Impact →Blast Radius · persona review
How it stays honest
The code graph is built from the AST, with no LLM. Risk is a weighted heuristic where every non-trivial factor is shown, not a model that predicts incidents. Remote MCP endpoint auth is Coming soon, so run the HTTP transport inside your network.
| Time | What happens |
|---|---|
| 09:10 | Your coding agent is about to rename customerId in a shared package. It asks Rigel first.find_consumers(package_names: ["@acme/orders-client"]) |
| 09:10 | → fulfillment-service · package · warning→ billing-worker · package · warningTwo services in other repos depend on the package. The agent keeps |
| 09:25 | The PR opens with a blast-radius score and persona review. |
| 13:00 | A design review asks how refunds reach billing. Ask Rigel answers with cited sources. You're not in the meeting. |
| 17:45 | The PR merges. Living Docs re-analyses only the affected docs and updates the rolling docs PR. |
Sample repos and times. Illustration.
Rigel builds a map of our services from code and keeps it current on every merge. Our coding agents can ask it which services in other repos depend on a package before they change it. Rigel's own agents (bug fixes, dependency upgrades, docs) only auto-merge when CI is green, reviews are resolved and the policy gates pass. Every PR can get a blast-radius risk score. It self-hosts in our cloud on our own model account (Beta).
Give your agent the map in your head.
Role · Product owners
From one-line ticket to signed-off requirement.
For product owners, business analysts and product managers.
The pain — Tickets arrive as one-liners, or as solution designs pretending to be requirements.
1
A requirement in business language.
Requirement Brief works one turn at a time into five fixed sections, plus open questions and risks. A critic sends back prescriptive technical language.
2
Start from what you already have.
Paste a brief, or point it at a Jira issue with its screenshots and PDF attachments.
3
Publish once, share with a link.
Immutable versioned snapshots, a share link and Markdown export. Then hand it straight to a grounded solution design.
Delivered by
How it works
The business-language critic blocks a turn when it spots implementation detail. It fails open, so a critic outage never blocks a product owner.
"Design the solution →" opens Solution Course from a finalised requirement. It reads services, the code graph and find_consumers, and an advisory grounding critic flags claims the chart doesn't support. Confluence push is Beta.
| Time | What happens |
|---|---|
| 09:00 | You paste a one-line Jira ticket, with a screenshot attached."Let members pause their plan" |
| 09:04 | Rigel asks the open questions, one turn at a time, and drafts the five sections live beside the chat. |
| 09:30 | You write "store it in a Kafka topic". The critic sends it back: outcomes, not implementation.prescriptive: names a technology |
| 10:15 | You publish a versioned snapshot and send the share link for sign-off. v1 publishedshare linkMarkdown |
| 11:00 | "Design the solution →". The draft names the three services involved, and flags the one claim the chart doesn't support. |
Sample ticket and times. Illustration.
Turn your next ticket into a requirement.
More roles
The rest of the team, briefly.
Release managers Rigel Impact →
The Daily Deploy Risk Digest on the schedule you set, and CAB packs built from the prod-to-staging diff.
For AWS
SRE and on-call Rigel Autopilot →Staging Alarm Triage walks a firing alarm, correlates it with the code graph and proposes a root cause. Open a read-only Ops Debug session and share it by link.
For AWS
Engineering managers and QA leads Rigel Autopilot →Jira bugs triaged by two investigators, a fix PR when they agree (a human merges it), and every agent PR in one Slack thread.
Early, on purpose
Become a design partner.
We're looking for a small group of engineering orgs that run 20+ services across many repos. Partners get early access, a direct line to the team that builds Rigel, and a say in the roadmap.