Start free. Pay when decisions go to production.
You can leave with your logic at any time.
The authoring environment and every published accelerator are free. You pay when a decision service runs in production, carrying its evidence, against your policy, on your infrastructure.
The full authoring environment, for learning, prototyping and non-commercial work.
- Full Rainbird Studio
- Every published accelerator, downloadable
- Run the evals locally
- Academy and community forum
- Non-commercial use
Production decisioning: one API and MCP endpoint, evidence attached to every call.
- Production REST API and MCP endpoint
- Foundry for agentic knowledge engineering
- Graded evals in your CI
- ModelOps: who changed what, when and why
- Named support and onboarding
For regulated estates with residency, assurance and policy-coverage obligations.
- Private cloud, self-hosted or in-region deployment
- SSO and role-based governance
- Contracted SLAs and security review support
- Computable Policy programme
- Customer Advisory Board seat
Sovereignty is a pricing term here, not a slogan: your rules, your graph, your file. You can leave with your logic at any time.
What buyers ask before they sign
What counts as a decision?
One call to a decision service that returns an outcome, a certainty and an evidence tree. Evals, tests and Studio authoring do not consume production decisions.
Can we self-host?
Yes, on Enterprise. Private cloud, in-region and fully self-hosted deployments are supported, along with data-residency commitments.
What happens to our knowledge graph if we leave?
You export it. The graph is a portable file, with its evals and its version history. Nothing about your decision logic is locked to us.
Is the Community tier really free?
Free forever for non-commercial use: the full authoring environment, every published accelerator and the evals. No credit card.
Prefer to see it work before a conversation? Download an accelerator and run its evals →
Start with the decision, not the contract
Take a validated accelerator, run its evals against your own cases, and only then talk to us about production.