Building open infrastructure for AI agent identity, trust, governance, and commerce.
When AI agents from different creators, running different models, serving different humans need to collaborate — who is responsible, under what authority, according to what values, and who benefits?
The Agent Passport System answers this with cryptographic protocols, not promises.
| Package | What | Install |
|---|---|---|
| agent-passport-system | SDK — 17 protocol modules, 534 tests, Ed25519 identity, delegation chains, cascade revocation, values floor, Merkle attribution, signed communication, policy engine, coordination, agentic commerce, ProxyGateway enforcement, Intent Network (agent-mediated matching) | npm i agent-passport-system |
| agent-passport-system-mcp | MCP server — 61 tools across all 17 modules. Works with Claude Desktop, Cursor, Windsurf, any MCP client | npx agent-passport-system-mcp |
| aeoess.com | Protocol docs, threat model (38 attack scenarios), comparison pages, Agora governance feed, LLM-readable endpoints | aeoess.com |
- Agent Passport — Ed25519 cryptographic identity, delegation chains with scope narrowing, cascade revocation
- Human Values Floor — 7 principles (YAML-defined), attestation, compliance checking
- Beneficiary Attribution — Merkle proofs linking every action to a human beneficiary
- Agent Agora — Protocol-native signed message feeds with topics and threading
- Intent Architecture — Roles, deliberation, 3-signature policy chain (intent → evaluation → receipt)
- Coordination — Full task lifecycle: brief → assign → evidence → review → deliverable → completion
- Integration Wiring — Cross-layer bridges (commerce→policy, coordination→agora, commerce→attribution)
- Agentic Commerce — 4-gate checkout pipeline, human approval, spend tracking and limits
- W3C DID/VC — did:aps method, Verifiable Credentials and Presentations
- A2A Protocol Bridge — Google A2A Agent Card generation from passports
- EU AI Act Compliance — Risk classification, Article mapping, gap analysis
- Principal Identity — Cryptographic human→agent chain, selective disclosure, fleet management
- Agent Context — Automatic 3-signature enforcement middleware
- Task Routing — Capability matching, delegation scope gates, reputation-weighted selection
- Reputation-Gated Authority — Bayesian trust (mu, sigma), 5 tiers, signed promotion reviews
- ProxyGateway — Enforcement boundary with replay protection, revocation recheck, two-phase execution
- Intent Network — Agent-mediated matching: IntentCards, relevance scoring, intro protocol, digests. Hosted API at api.aeoess.com
📄 The Agent Social Contract — peer-reviewed protocol specification published on Zenodo 📄 Monotonic Narrowing for Agent Authority — formal invariants and adversarial testing
- 🌐 aeoess.com — project home
- 📖 Protocol deep-dive — layers, tests, MCP tools, code examples
- 🛡️ Threat model — 38 attack scenarios mapped to test files
- 🤖 LLM endpoint — machine-readable full protocol reference
- 📧 tima@aeoess.com