# Certior

> A capability boundary for multi-agent AI. Every tool call is checked against a Lean-proven policy before it runs.

## Docs

- [Authentication](https://docs.certior.io/api/authentication.md): Register a user, log in, rotate an API key, read the current user. API keys go in the Authorization header.
- [Compliance](https://docs.certior.io/api/compliance.md): List built-in policy presets. Export the audit package for a finished execution.
- [Executions](https://docs.certior.io/api/executions.md): List executions, fetch one by id, cancel a queued or running execution.
- [REST API overview](https://docs.certior.io/api/overview.md): The Certior REST API covers verified task execution, workflow orchestration, compliance export, releases, and trust badges under the /api/v1/ prefix.
- [Releases](https://docs.certior.io/api/releases.md): Release decisions, snapshot promotions, GitHub webhook, and the trust badge endpoint.
- [Tasks](https://docs.certior.io/api/tasks.md): Submit a task for verified execution. Certior runs the three gates before each tool call and returns an execution id.
- [Workflows](https://docs.certior.io/api/workflows.md): Multi-stage orchestrations where each stage runs under its own policy and binds to a previous stage by stage id.
- [Capability model](https://docs.certior.io/concepts/capability-model.md): Capabilities are permission strings. A guard's permissions are its ceiling. A child guard's set must be a subset of its parent's. Z3 enforces, Lean proves the rule.
- [Certificates](https://docs.certior.io/concepts/certificates.md): Every allowed call produces a VerifiedCertificate carrying the verified properties, the proof trace, the plan hash, and a signature.
- [How it works](https://docs.certior.io/concepts/how-it-works.md): Three gates run before every tool call: capability, content, budget. Z3 enforces, Lean proves the model is sound.
- [Compliance policies](https://docs.certior.io/concepts/policies.md): Four built-in presets - default, hipaa, sox, legal_privilege - bundle content rules, permission ceilings, audit retention, and proof obligations.
- [CrewAI](https://docs.certior.io/guides/crewai.md): Wrap one CrewAI tool with certior_tool_wrapper, or guard every tool in an existing crew with guard_crew_tools.
- [Custom orchestration loop](https://docs.certior.io/guides/custom-loop.md): Use Guard.verify() or Guard.wrap() to add capability, content, and budget gates to any agent loop you already own.
- [LangChain](https://docs.certior.io/guides/langchain.md): Attach CertiorCallbackHandler to any LangChain chain or agent. Every tool call is verified before execution.
- [OpenAI tool calling](https://docs.certior.io/guides/openai.md): Drop verify_tool_calls() onto an OpenAI tool-calling response. Native shape, no transform.
- [OpenClaw](https://docs.certior.io/guides/openclaw.md): GuardedAgent enforces, CertiorCallbackHandler accounts, GuardedPipeline delegates with subset checks. Addresses every threat in arXiv:2603.12644.
- [GitHub Action](https://docs.certior.io/integrations/github-action.md): Gate every pull request with Certior policy checks. The action evaluates declared capability surfaces against the configured policy package.
- [Skill audit CLI](https://docs.certior.io/integrations/skill-audit.md): certior-skill-audit proves a skill's declared capability surface is a subset of a parent guard's permissions before the skill is wired in.
- [Certior Studio](https://docs.certior.io/integrations/studio.md): An optional Next.js UI for the Certior FastAPI server. Inspect runs, explore the delegation graph, export audit packages.
- [Certior](https://docs.certior.io/introduction.md): A capability boundary for multi-agent AI - every tool call is checked against a Lean-proven policy before it runs.
- [Quickstart](https://docs.certior.io/quickstart.md): Install Certior, create a Guard, see an allowed call return a certificate and a blocked call raise CertiorBlocked.
- [Configuration](https://docs.certior.io/reference/configuration.md): Guard constructor parameters, environment variables, and the one env var you set to enable live Lean flow verification.
- [Live Lean verification](https://docs.certior.io/reference/lean-binary.md): Build certior-flow-check to add live per-plan flow verification on top of Certior's always-on Z3 gate.
- [Trust package](https://docs.certior.io/reference/trust-package.md): What Certior formally proves, what it does not claim to prove, and how runtime evidence binds to those proofs.

## Optional

- [GitHub](https://github.com/paulinebourigault/certior)
- [PyPI](https://pypi.org/project/certior/)