Build, configure, and integrate CLAiRE.
Two starting points. Configuration covers the building blocks you assemble inside CLAiRE — agents, skills, flows, ontologies, and more. Integration covers the wire protocols other systems use to talk to CLAiRE, including the REST API reference (OpenAPI) and the GraphQL surface.
The building blocks you assemble.
CLAiRE is composed of a small set of orthogonal concepts. Master these and you can model any quality workflow without writing custom code.
Agents
Reasoning units with a goal, a model, a toolbox, and guardrails. Configure their persona, tools, and guardrail policies.
Skills
Composable capabilities an agent can invoke — tool calls, prompt fragments, and procedure references bundled with input/output contracts.
Flows
Directed graphs of nodes and transitions that orchestrate agents, human approvals, and system actions across a record's lifecycle.
Ontologies
The shared vocabulary that grounds every AI call — record types, fields, controlled values, and the relationships between them.
Guardrails
Four-layer policy enforcement — input filters, tool constraints, data redaction, and output post-filters — applied to every AI call.
Evidence & Audit
Every action in CLAiRE is recorded in a tamper-evident, hash-chained audit trail. Configure retention, classification, and signing.
How other systems talk to CLAiRE.
CLAiRE exposes two API surfaces. The bulk of the application is GraphQL; a small set of HTTP/REST endpoints handles health checks, SSO callbacks, and the Model Context Protocol transport.
REST API reference
Health, MCP transport, and SAML/SSO callbacks. Includes the full OpenAPI 3.1 specification with a downloadable openapi.yaml.
GraphQL endpoint
The primary application API. All record CRUD, agent invocation, flow execution, and audit queries run through /graphql.