Agent Passport

Enforcement and accountability layer for AI agents. Bring your own identity (did:key, did:web, SPIFFE, OAuth, did:aps). Gateway enforcement boundary, monotonic narrowing, cascade revocation, spending controls, data lifecycle, observation governance (telemetry scopes, derivation rights, behavioral memory). Use when agents need scoped delegation, trust scoring, constraint enforcement, or cryptographic audit trails. SDK leads with the /core subpath (24 curated functions), MCP leads with APS_PROFILE=essential (25 tools covering identity, delegation, enforcement, commerce, reputation). 5,281 tests. 8 framework adapters: Stripe, Composio, IBAC/Cedar, LangChain, CrewAI, MCP, A2A, Gonka. Full surface area (107 modules, 152 MCP tools) still available under APS_PROFILE=full and the root import. SDK 6.0.1 and MCP 6.0.1 are current on npm, Python 3.0.1 on PyPI, Rust 0.3.0 on crates.io, Go v0.7.0. The 3.3.1 release added Delegated Action Evidence: bilateral pair reconciliation with five mismatch classes, verifier-side revocation observation receipts under a stated freshness contract, Merkle-rooted evidence bundles with a per-axis claim-state report (verify-bundle), spec 4.1 action_ref canonicalization validated by cross-language vectors, and jurisdiction selection records that surface pack conflicts, on top of evidentiary type safety, Wave 1 accountability, Instruction Provenance Receipt, and bilateral receipts.

Agent Passport System

When to use this skill

• Agent needs cryptographic identity (Ed25519 passport)
• Delegate authority between agents with scope, spend limits, depth controls
• Revoke access: one call kills all downstream delegations
• Run agent commerce with 5-gate checkout (passport, delegation, merchant, spend)
• Coordinate multi-agent tasks (assign, evidence, review, deliver)
• Track data contributions with Merkle proofs
• Encrypt agent-to-agent communication (E2E, forward secrecy)
• Score agent trust (Bayesian reputation, passport grades 0-3)
• Enforce values compliance (8 principles, graduated enforcement)
• Found institutions with charters, offices, approval policies

Install

bash

npm install agent-passport-system        # SDK: /core subpath is the curated default
npm install agent-passport-system-mcp    # MCP server: APS_PROFILE=essential is the default
go get github.com/aeoess/agent-passport-go@v0.7.0   # Go SDK, byte-parity subset (passport, delegation, attribution, completion, in-toto, values)
pip install agent-passport-system==3.0.1             # Python SDK
cargo add agent-passport-system@0.3.0                # Rust SDK, library crate agent_passport

Minimal SDK import (lead with the curated essentials):

typescript

import {
  createPassport, createDelegation,
  evaluateIntent, commercePreflight, generateKeyPair
} from 'agent-passport-system/core'

Minimal MCP install (essential profile is the default; APS_PROFILE=full for all 152 tools):

bash

npx agent-passport-system-mcp

Remote MCP (zero install): https://mcp.aeoess.com/sse

Core workflow

1. Create identity → returns passport + keypair

bash

npx agent-passport join --name my-agent --owner alice

Output: .passport/agent.json with Ed25519 keypair, signed passport, values attestation. Treat like an SSH key.

2. Delegate authority → returns signed delegation

bash

npx agent-passport delegate --to <publicKey> --scope web_search,commerce --limit 500 --depth 1 --hours 24

Output: signed delegation with scope, spend limit, max depth, expiry. Authority can only narrow at each transfer.

3. Record work → returns signed receipt

bash

npx agent-passport work --scope web_search --type research --result success --summary "Found 3 sources"

Output: Ed25519-signed receipt traceable to a human through the delegation chain.

4. Prove contributions → returns Merkle proof

bash

npx agent-passport prove --beneficiary alice

Output: Merkle root + inclusion proofs. 100K receipts provable with ~17 hashes.

MCP tools (152 total)

Setup: npx agent-passport-system-mcp setup (auto-configures Claude Desktop + Cursor)

Identity & trust (11 tools): generate_keys, identify, issue_passport, verify_issuer, create_principal, endorse_agent, get_passport_grade, list_issuance_records, get_behavioral_sequence, verify_endorsement, revoke_endorsement

Delegation & revocation (5): create_delegation, verify_delegation, revoke_delegation, sub_delegate, create_v2_delegation

Commerce & wallets (4): commerce_preflight, get_commerce_spend, request_human_approval, create_checkout

Coordination (11): create_task_brief, assign_agent, accept_assignment, submit_evidence, review_evidence, handoff_evidence, get_evidence, submit_deliverable, complete_task, get_my_role, get_task_detail

Communication (7): send_message, check_messages, broadcast, list_agents, post_agora_message, register_agora_agent, register_agora_public

Governance & policy (12): load_values_floor, attest_to_floor, create_intent, evaluate_intent, create_policy_context, create_agent_context, execute_with_context, create_charter, sign_charter, verify_charter, create_approval_request, add_approval_signature

Data attribution (10): register_data_source, create_access_receipt, create_derivation_receipt, create_decision_lineage_receipt, record_training_use, check_data_access, check_purpose_permitted, check_retention_expired, query_contributions, generate_compliance_report

Intent Network (5): publish_intent_card, remove_intent_card, search_matches, request_intro, respond_to_intro

Framework adapters (8)

One-function governance for every major agent framework. Each wraps tool/task execution with APS delegation checks and Ed25519-signed receipts.

typescript

import {
  governLangChainTool,              // LangChain/LangGraph
  governCrewTask,                   // CrewAI
  governMCPToolCall,                // Any MCP server
  governIBACIntent,                 // IBAC (Cedar/OPA)
  passportToA2ACard,                // A2A Agent Cards
} from 'agent-passport-system'

// Also available as standalone packages:
// npm install @aeoess/stripe-governance
// npm install @aeoess/composio-governance

Adapter	Function	What it wraps
LangChain	governLangChainTool()	BaseTool.invoke()
CrewAI	governCrewTask()	Crew task execution
MCP	governMCPToolCall()	Any MCP tool call
IBAC/Cedar	governIBACIntent()	Cedar/OPA policy tuples
A2A	passportToA2ACard()	Agent Card ↔ passport bridge
Stripe	governMPPPayment()	Stripe agent payments
Composio	governComposioAction()	catalog of 250+ tool targets (subject to the upstream Composio registry and local configuration)
Gonka	governGonkaInference()	Decentralized GPU compute

Programmatic API

typescript

import {
  joinSocialContract,   // → { passport, keyPair, attestation }
  createDelegation,     // → signed Delegation
  processToolCall,      // → { permitted, constraintResults, receipt }
  cascadeRevoke,        // → { revoked: string[], receipts }
  computePassportGrade, // → 0 | 1 | 2 | 3
  createIssuanceContext, // → IssuanceContext with evidence + assessment
} from 'agent-passport-system'

Composition-check receipt (CompositionCheckV0)

A chain of individually rule-legal delegations can compose to a globally-unsafe target that per-hop monotonic narrowing cannot detect (each delegation narrows scope correctly, but the composed chain routes to an unsafe aggregate). The SDK carries proof that an external attestor ran a composition-hazard check; detection of the hazards is private gateway intelligence and is not in the SDK.

typescript

import { CompositionCheckV0 } from 'agent-passport-system'

const result = CompositionCheckV0.verifyCompositionCheck(receipt, context)
// verifies the anchor: signature, binding to (chain_hash, action_ref, context_hash),
// freshness, attestor trust. Surfaces independence_is_second_anchor corroborated from the
// caller's trust context, never the receipt's self-declaration.

The verifier checks the anchor, not the composition. It evaluates no policy and emits no aggregate safe verdict: a per-check pass means only that the named attestor reported pass for the named profile over the bound context, never global safety. gateway_self is always weak (one trust domain); only a context-corroborated independent_registered attestor is a second anchor, mirroring how RAP-v0 gates its strong claim on domains >= 2. Conformance vectors in conformance/composition-check/v0/.

traceBeneficiary().verified is a real cryptographic check: true only when the receipt signature verifies at the chain tail and every delegation hop passes verifyDelegation. The separate resolved field carries lookup-success semantics (the lineage maps to known records) without making a cryptographic claim.

Passport grades (attestation architecture)

Grade	Meaning	Trust signal
0	Bare Ed25519 keypair	Unverified
1	Issuer countersigned	AEOESS processed
2	Runtime-bound + challenge-response	Infrastructure-attested
3	Runtime + verified human principal	Full chain of trust

Grade travels with the passport. Any consumer reads it without understanding scoring internals.

Key facts

• Enforcement and accountability layer: bring your own identity, gateway does the rest
• SDK /core subpath: 24 curated functions for 90% of integrations
• MCP essential profile: 25 tools by default (identity, delegation, enforcement, commerce, reputation)
• Policy eval <2ms, 403 ops/sec, 14 constraint dimensions
• 5,281 tests including 38 adversarial scenarios. These are a developer-authored internal evaluation, not an independent red-team or neutral third-party adversarial harness.
• Zero heavy dependencies: Node.js crypto + uuid only
• Apache-2.0 license
• Full surface area: 107 modules, 152 MCP tools, available under APS_PROFILE=full and the root agent-passport-system import.

Links

• npm: https://www.npmjs.com/package/agent-passport-system
• MCP: https://www.npmjs.com/package/agent-passport-system-mcp
• PyPI: https://pypi.org/project/agent-passport-system/
• crates.io: https://crates.io/crates/agent-passport-system
• Go: https://pkg.go.dev/github.com/aeoess/agent-passport-go
• GitHub: https://github.com/aeoess/agent-passport-system
• Docs: https://agent-passport.org/llms-full.txt
• Paper: https://doi.org/10.5281/zenodo.18749779