Ratify Labs
Explore working demonstrations and open technical references that show how delegated authority survives agent frameworks, runtime boundaries, and external receivers.
Reference catalog
Status labels distinguish live labs from tested references and work still in development.
Interactive authority lab
● LiveRun signed Ratify proofs through the published TypeScript verifier and try to exercise authority the principal never granted.
What you can prove
See the exact protocol verdict, receiver policy decision, and measured verification time for every attempt.
Verified: August 2026
Executable reference
◆ Tested referenceA LangChain agent carries principal-signed authority through a public MCP tool interceptor to an independently started receiver.
What you can prove
Verify that the protected handler runs once for a valid action and remains untouched across adversarial denials.
Verified: 24/24 cases · August 2026
Executable reference
◆ Reference under reviewA Google ADK workflow presents bounded Ratify authority to a receiver that verifies before executing the requested action.
What you can prove
Inspect a framework-specific adapter while preserving receiver-owned verification and policy.
Executable reference
◆ Reference under reviewA model-agnostic agent carries resource-bound authority through MCP and OpenShell to an independent receiver.
What you can prove
Explore how runtime controls and portable delegated authority compose without replacing one another.
Phase 1 build
◌ In developmentA LangChain agent in an isolated Maritime runtime requests a bounded action from a separately deployed MCP receiver.
What you can prove
Watch the same recognized agent succeed within its delegation and fail before execution when it exceeds that authority.
Bring your boundary
Work with Ratify to turn one consequential action into a bounded, adversarially tested technical reference.
