Control tool use
Decide whether an agent may call a tool or perform an operation before execution.
Cedarling is an embeddable authorization engine for applications and AI agents. It evaluates Cedar policies close to the action using either JWTs from trusted issuers or identity and context established by your application, then returns an ALLOW or DENY decision with diagnostics.

Cedarling lets your application check whether an agent may use a tool, read data, or change a workflow before execution.

Decide whether an agent may call a tool or perform an operation before execution.
Apply Cedar policies across agents, users, services, and environments.
Embed Cedarling in-process or place it beside the service that owns the protected action.
Can this principal perform this action on this resource in this context?
In token-based authorization, Cedarling validates one or more mapped JWTs and makes their claims available to Cedar policies.
In application-asserted authorization, your application supplies the principal and is responsible for establishing its identity.
Example
Example
Example
Example
Use Cedarling in an application process or alongside it, from browser and mobile experiences to backend services, API gateways, databases, and sidecars.

Bring Cedarling into JavaScript, Python, Rust, Go, Java, Kotlin, Swift, C, mobile applications, gateways, databases, or a language-independent sidecar.
Open a reviewed policy store, change a safe request, and run Cedarling locally in your browser.
Most Playground samples call authorizeUnsigned with application-asserted identity. Google Trusted Identity calls authorizeMultiIssuer with an ID token from a configured trusted issuer.