One identity. Every version.
A stable passport for your agent, even as its model, framework, or infrastructure changes.
An identity that moves with your agent.
Connect deployments. Prove control.
Stay recognizable across every runtime.
YOUR RUNTIME.
YOUR RULES.
Models change. Frameworks evolve.
Your agent’s identity should endure.
One passport connects it all.
A stable passport for your agent, even as its model, framework, or infrastructure changes.
Link each runtime with its own signing key. Connect your existing agents on your terms.
Authorize a deployment. Revoke a compromised key. Keep the rest of your agent running.
An identity is only useful if you can verify it.
Try a real signature check in your browser,
then see what happens when trust changes.
ap:near:testnet:42
Select a deployment and run a check. The evidence will speak for itself.
A small integration. A persistent identity.
Designed to fit the agent you’ve already built.
Establish an owner-controlled identity on NEAR.
Generate a local key. Authorize its public half.
Make your agent’s identity independently checkable.
import { AgentPassport } from "@agentpassport/sdk";// Your agent. Its own identity.const passport = new AgentPassport({ passportId: "ap:near:testnet:42", deploymentId: "deployment-001", privateKey: process.env.AGENT_SIGNING_KEY});// Sign locally. Share only what you choose.await passport.record({ type: "task.completed", taskId: "research-001", outcome: "success"});A persistent identity for a logical agent. It connects an owner account to separately authorized deployment keys, so the same agent can be recognized across runtimes. AgentPassport Connect is the proposed protocol behind whoami.
The protocol is designed to work alongside your existing agent. A framework-independent TypeScript SDK is planned first, followed by an OpenAI Agents SDK adapter. Your prompts, models, and hosting stay under your control.
Verification establishes a signature’s validity and a key’s authorization. It does not prove that an agent is safe, that its advertised capabilities work, or that a self-reported task actually happened. Independent evaluation credentials are a future milestone.
The proposed NEAR registry records ownership, deployment authorization, revocation, and public metadata commitments. Agent execution and activity receipts stay offchain. Private keys, prompts, and conversations are never part of the registry.
You can explore this frontend and run the local cryptographic playground now. The registry, published SDK, wallet connection, and hosted verification service are still in development. The playground makes no blockchain transactions.
One agent. One passport. Endless possibilities.