Platform fee
The chart, the agents, the gates and the record.
RigelPricing
Design-partner pricing · Cloud plans published before launch
A platform fee. Your model bill stays yours.
You pay Rigel for the platform. Model usage runs on your own Anthropic key or Claude through your AWS account, and embeddings on your own Voyage AI key. Every token shows up on your own invoices, at your own rates.
Rigel Cloud plans Coming soon · Rigel Enterprise self-host through sales today Beta
The model
Where your money goes.
Two bills, kept apart on purpose.
Your team
Uses Rigel
- maps and refreshes
- agents and gates
- chat and drafting
Model usage
Your Anthropic key, or Claude through your AWS account. Capped per run.
Embeddings
On your own Voyage AI key, for search and reranking.
Illustration. Self-host routes all model usage to your accounts today. Your own Anthropic key in Rigel Cloud is Coming soon.
Prerequisite for code-change auto-merge: GitHub Copilot code review.
1
Platform fee
On design-partner terms today, for the chart, the agents, the gates and the record.
2
Your model accounts
Analysis, agents and chat run on your Anthropic key or Claude through your AWS account, plus your Voyage AI key for embeddings.
3
Budgets you set
Every run, chat query and drafting session has a USD cap, so no agent loop runs away. Organisation-level caps are Coming soon.
4
Copilot seats, if needed
GitHub Copilot seats, if you want code-change PRs from Rigel's agents to auto-merge.
The full bill
Everything you'd pay for.
Five lines, and who sends each invoice.
The full bill.
1
Rigel platform fee — on design-partner terms.
2
Model usage — your Anthropic key or Claude through your AWS account. Every run has a USD cap; organisation-level caps Coming soon.
3
Embeddings — your Voyage AI key. Code excerpts are sent to Voyage.
4
GitHub Copilot code review — needed only if you want Rigel's code-change PRs to auto-merge. Without it they wait for a person. Pluggable reviewers are on the roadmap.
5
Your infrastructure — Postgres and the hosts you run Rigel on.
What drives your model bill
Onboarding
A one-time Deep Survey per repo: extraction plus seven analysis passes per service. Cached per commit, so re-runs on unchanged code cost nothing extra.
Every merge
A triage step (capped at $2) plus re-analysis of only the affected cells.
Every PR with Impact on
The risk score uses no model; the persona reviews do.
Agent runs (fixes, upgrades, triage)
Each has its own USD budget and turn cap.
Because model and embedding usage runs on your own accounts, you see the actual spend on your own bill — not a marked-up line from us. We'll walk through expected usage for your estate on a call.
Pricing
How pricing works today.
Rigel Enterprise (self-hosted, Beta) is sold directly, on design-partner terms we agree on a call. You pay Rigel a platform fee. Model and embedding usage runs on your own accounts, so it never shows up marked up on our invoice.
We set the platform fee on a call, around the value metric that fits your estate — per repo, per service or per active developer. We're not publishing a list price while Enterprise is in design-partner Beta.
Language support today: Node.js and TypeScript estates; Java, Python and Go are Coming soon. GitHub only today.
Rigel Enterprise · self-host Beta
Map, MCP, Impact, Autopilot and Plan; single sign-on with your OpenID Connect provider; your own Anthropic key or Claude through your AWS account.
Talk to us about design-partner terms →Rigel Cloud Coming soon
Plans and prices will be published before it opens.
Maturity
What's shipped today.
What you'd pay for, labelled honestly.
| Status | What you'd pay for |
|---|---|
| Shipped | Rigel Map, MCP (20 read-only tools), Impact, Autopilot, Plan. |
| Beta | Self-host, endpoint and queue consumer matching, Confluence publishing. |
| Coming soon | Rigel Cloud, Java/Python/Go, supervising any agent's PRs, remote MCP endpoint auth. |
Cost control
Know what agents cost. Cap it.
Budget caps
On every run, chat query and drafting session. A run that hits its cap stops.
Cost on every run
Tokens and estimated cost on each automation run and chat session.
Admin usage view
Adoption, quality signals and cost across chat, requirements and solution design.
Fix loops capped at 4
A stubborn CI failure can't burn budget indefinitely.
Cached extraction
Re-running analysis on an unchanged commit reuses earlier results.
Org-level spend caps Coming soon
Today many capped runs can still add up. Watch the admin view.
FAQ
Questions, answered.
What does "bring your own key" mean?
Model calls go to your own provider account, so the spend lands on your bill at your rates. In a self-hosted install that's true today, on your Anthropic key or Claude through your AWS account. When Rigel Cloud launches, you'll add your own Anthropic key, stored encrypted, and Rigel won't fall back to a platform key. Coming soon
Where does my code go?
Rigel reads your repos to build the chart. Code excerpts and derived docs go to your model provider (the Anthropic API, or Claude through your AWS account) and to Voyage AI for embeddings and reranking.
Which models does Rigel use?
Claude models (Opus, Sonnet and Haiku), through the Anthropic API or your AWS account, selectable per deployment. Embeddings use Voyage AI.
Can we self-host?
Yes. Rigel Enterprise runs in your own cloud today, in Beta, on Docker Compose, Kubernetes or an AWS reference stack.
When does Rigel Cloud open?
Rigel Cloud is Coming soon. We haven't set a date yet. Prices will be published before it opens, and the waitlist hears first.
Is there a design partner program?
Yes, for engineering orgs with 20+ services across many repos. Partners get early access, a direct line to the team and input on the roadmap. Commercial terms are agreed on a call.
How long does rollout take?
Start with one application's repos. See what rollout takes →
Which languages and platforms?
Node.js and TypeScript estates get the full chart today. Java, Python and Go service discovery is coming soon. Dependency upgrades cover the npm ecosystem today. GitHub only today.
Does Rigel replace our reviewer or agents?
No. Rigel briefs your agents with the chart and can score their PRs. Its own agents' PRs pass its policy gates, and for code changes it requires GitHub Copilot code review as the reviewer of record.
Can you support our security review?
Yes. Request the security pack. Rigel holds no compliance certifications today, and we'll walk your team through each mechanism.
Start with your own repos.
Design-partner pricing is set on a call. Cloud plans will be published before Rigel Cloud opens.
Early access · A direct line to the team · Input on the roadmap