Every agent gets a cryptographic DID. Sign messages, verify sources, and discover agents by capability — with three lines of code.
from agentid import Agent # One call — instant cryptographic identity agent = Agent.create( name="research-agent", capabilities=["web-search", "summarization"], registry_url="https://api.agentid-protocol.com", ) # → did:agentid:7sP3V2mNkQxRtYbLcDfHgJwAeUiMoZnXvBqKpTsWyE print(agent.did) # Sign outputs — anyone can verify signed = agent.sign({"result": "task complete"}) Agent.verify_from_did(signed) # → True
Every team building with AI agents hits the same three walls. AgentID tears them down.
No standard identity — every team hand-rolls their own auth. Agents can't prove who they are across systems.
No discovery — agents are hardcoded or manually configured. Finding the right agent for a task is manual work.
No signing — outputs can be spoofed or tampered with. There's no way to verify a message is authentic.
Every agent gets a DID derived from an Ed25519 keypair. No sign-up. Works offline. Cryptographically verifiable.
Ed25519 key generated locally. The DID is derived from the public key — no server involved.
Submit your DID, capabilities, and a cryptographic proof. The registry verifies ownership before accepting.
Sign messages with your private key. Anyone verifies using only the DID — no shared secrets, no central login.
from agentid import Agent # Create — generates keypair, registers agent = Agent.create( name="analyst-v2", capabilities=["analysis", "reporting"], owner="team@company.com", registry_url="https://api.agentid-protocol.com", ) # Discover agents by capability agents = Agent.find( capability="analysis", registry_url="https://api.agentid-protocol.com", ) # Sign a message signed = agent.sign({"output": "Q3 revenue up 18%"}) # Verify from anywhere — no shared secret valid = await Agent.verify_from_did( signed, registry_url="https://api.agentid-protocol.com", ) print(valid) # True
Agent-to-agent authentication in 4 lines
from agentid import Agent agent = Agent.create( name="worker-agent", capabilities=["analysis"], ) # → did:agentid:5Ht...
Agent generates an Ed25519 keypair locally and publishes a cryptographic DID to the registry. No central authority required.
signed = agent.sign({ "task": "analyze dataset", "priority": "high", }) # Attaches DID + signature
The worker agent signs its message with its private key. The resulting envelope carries the DID, a nonce, and a timestamp alongside the payload.
valid = Agent.verify_from_did( signed, ) print(valid) # True
Any peer agent resolves the DID from the registry and checks the signature — no shared secrets, no API keys, no manual config.
{ "payload": { "task": "analyze dataset", "priority": "high" }, "signature": "MEQCIBx7k2nP...3dQsYaR", "signer_did": "did:agentid:5HtNkQxRtYbLcDfHgJw", "timestamp": 1714234567, "nonce": "a3f9c12e" }
Drop-in identity for the leading AI agent frameworks. One install, full AgentID support.
Callback handler, identity tools, and automatic signing for LangChain agents and chains.
pip install agentid-protocolConversableAgent mixin with built-in discovery and verification for multi-agent conversations.
pip install agentid-protocolAgent wrapper with AgentID tools for crew-to-crew trust and capability-based discovery.
pip install agentid-protocolFull-featured SDK for TypeScript and Node.js projects. Identical API to the Python SDK.
npm install @vikhulus/agentid-protocolThe protocol is open source and free forever. Pay for the hosted registry and enterprise features.
Open source, cryptographically sound, and ready for production.