Securing the
Agentic Future.
We build the safety and security foundation for AI agents — identity, policy, verification, and runtime trust.
ThirdKey is an AI safety and security company. As autonomous agents move from demos into production, the systems around them — how they’re identified, what tools they can call, which schemas they trust, how policy is enforced — become critical infrastructure. Symbiont, SchemaPin, AgentPin, and ToolClad are open, interoperable primitives for deploying agents you can actually trust.
Same agent. Secure runtime.
A Rust-native, policy-governed runtime for executing AI agents and tools in production. Cedar authorization, cryptographic tool verification, sandboxed execution, and tamper-evident audit — for agents you can actually deploy with sensitive data and real risk.
Learn More →SchemaPin 🧷
Cryptographic schema verification for AI agents and MCP servers. Prevent “MCP Rug Pull” attacks with ECDSA signatures, .well-known DNS-anchored trust, and TOFU key pinning.
AgentPin 🪪
Domain-anchored cryptographic identity for AI agents. Issue short-lived ES256 credentials, verify with a 12-step protocol, and enforce capability-scoped access — all anchored to your domain.
Learn More →ToolClad 🔧
Declarative tool interface contracts for agentic runtimes. A single .clad.toml manifest defines typed parameters, validation, invocation, output parsing, and Cedar policy — across oneshot CLI, interactive sessions, and governed browser execution.