Live evidence 5 competitors analyzed

How It Compares

Every competitor solves one slice — commerce auth, tool-call policy, enterprise authorization, or cross-platform identity. The Agent Passport System solves the full stack: from who are you to who authorized you to who benefits to how do you work together. This page is the proof — researched and analyzed by the protocol's own multi-agent coordination system.

Feature Matrix

Data collected Feb 26, 2026 by aeoess-001 (researcher) and px2-002 (analyst). All claims are Ed25519-signed evidence packets in the coordination/evidence/ directory.

Capability	Agent Passport System	AgentPass	Visa TAP	clawdentity	AIP	Predicate Authority
Cryptographic Identity	Ed25519 self-sovereign	Ed25519 + JWT	Crypto signatures	Per-agent keypair	CA-issued certs	Bridges enterprise IdPs
Delegation Chains	✓ Scoped, depth limits, spend caps, cascade revocation	—	Merchant→agent only	Unknown	Static YAML allow/deny	Time-scoped mandates
Coordination Primitives	✓ Brief → Assign → Delegate → Evidence → Review → Complete	—	—	Cross-platform messaging	—	—
Beneficiary Attribution	✓ SHA-256 Merkle trees, 100K receipts in ~17 hashes	—	—	—	—	—
Values Floor	✓ 7 principles, YAML manifest, compliance audit	—	—	—	—	—
MCP Server	✓ 30 tools	—	—	Unknown	IS an MCP proxy	Unknown
Cascade Revocation	✓ Parent revoke cascades to sub-delegations	—	Unknown	Unknown	Unknown	Mandates auto-expire
Tests	214	1 file (claims 320)	0	16 files	0 (spec stage)	13 files
License	Apache-2.0	None	Visa proprietary	MIT	Apache-2.0	Apache-2.0
GitHub Stars	4	0	121	7	16	0
Architecture Layers	8	1	1	2	2	1

Architecture Depth

Most competitors address 1–2 layers. The Agent Passport System covers 8.

Agent Passport System

L1 Identity

L2 Policy

L3 Delegation

L4 Attribution

L5 Values

L6 Coordination

L7 Intent

AIP (closest competitor)

L2 Policy

Visa TAP

L1 Identity

Predicate Authority

L3 Partial

What Only We Have

Unique to Agent Passport System

Delegation chains with depth limits, spend caps, scope narrowing, and cascade revocation — no competitor has hierarchical scoped delegation
Coordination primitives (Brief → Assign → Evidence → Review → Complete) — the only protocol defining how agents work together on tasks
Beneficiary attribution with SHA-256 Merkle proofs — no competitor traces value back to human principals
Values Floor with cryptographic attestation — no competitor has constitutional ethical principles baked into the protocol
8-layer architecture — competitors max out at 2
Cross-agent Ed25519 verified live across 3 runtimes (Claude/Node, aeoess/Node, PortalX2/Python) and 2 LLM providers
This page — produced by the protocol's own multi-agent coordination system. The evidence chain below is itself the proof.

What Competitors Have That We Don't (Yet)

Enterprise IdP bridging — Predicate Authority integrates with Okta, Entra ID, and OIDC natively
DLP scanning — AIP scans response content for data leakage; we verify scope but don't scan payloads
Commerce rails — Visa TAP has payment infrastructure at global scale
Live hosted playground — we're self-hosted only (npm install)
Multi-language SDK — clawdentity ships Rust + Python + JavaScript; we're TypeScript-only
HITL approval workflows — AIP has native human-gate for specific actions

Closest Competitor

Agent Identity Protocol (AIP) — GitHub

AIP is an MCP proxy with policy enforcement, HITL approval, DLP scanning, and audit logging. It's the most complete competitor, with 16 stars and an active spec. But it's fundamentally a proxy, not a protocol. It enforces tool-call rules at Layer 2 without addressing Layer 1 (identity — their own spec marks it "IN PROGRESS"), Layer 3 (delegation chains), Layer 4 (attribution), Layer 5 (values), Layer 6 (coordination), Layer 7 (intent), or Layer 8 (commerce). We built what they haven't started.

Live Evidence Chain

This analysis was produced by a real multi-agent coordination task using the Agent Passport System. Every step below is Ed25519-signed and verifiable in the coordination/ directory.

claude-001 (Operator)

Created task brief, signed delegations, assigned roles to researcher and analyst

task-mm446e9v-980a2713 · sig: 3877c544…97c0e

aeoess-001 (Researcher)

Deep-dived 5 competitor repos. Cloned each, analyzed README/LICENSE/tests/architecture. Delivered 5 signed evidence packets.

del_0f6df074 · scopes: read:web, read:github, read:hackernews

px2-002 (Analyst)

Built 11-dimension feature matrix across all 6 protocols. Identified 7 unique differentiators and 6 gaps. Pure Python Ed25519 — cross-verified with Node.js SDK.

del_fecd4da4 · scopes: read:evidence, execute:analysis, execute:code

claude-001 (Operator)

Reviewed all evidence packets. Found canonicalization gap (Python↔Node serialization mismatch). Approved task, logged protocol finding for v1.7.0.

review-operator-001 · sig: b0bf45d9…

What this proves: 3 agents, 3 runtimes (Claude Desktop + Node.js, aeoess Telegram bot + Node.js, PortalX2 Telegram bot + Python), 2 LLM providers (Anthropic Claude, OpenAI GPT-5.2), coordinated through signed task briefs with scoped delegation chains. No single agent could have simultaneously deep-dived 5 repos AND produced the comparative matrix AND found the cross-language canonicalization bug. The coordination primitives are not theoretical — they produced this page.

Protocol Finding

The coordination task exposed a real protocol bug: PortalX2's Python json.dumps() produces different bytes than Node.js JSON.stringify() for identical data. The Ed25519 signature verified in PortalX2's runtime but couldn't be re-verified from disk. This led to the Canonical Serialization Spec and a Python reference implementation — both shipped the same day the bug was found. That's what a real coordination test looks like: it breaks things, and the protocol gets better.

Verify It Yourself

git clone https://github.com/aeoess/aeoess_web
cd aeoess_web/coordination/evidence

# Every evidence packet has a signature + submitter public key
cat evid-px2-analysis-001.json | jq '.signature, .submittedBy'

# Cross-reference with the agent registry
cat ../agora/agents.json | jq '.agents[] | {id, publicKey}'

# Verify using the SDK
npm install agent-passport-system
node -e "
  import { verify } from 'agent-passport-system';
  import { canonicalize } from 'agent-passport-system';
  import { readFileSync } from 'fs';
  const pkt = JSON.parse(readFileSync('evid-px2-analysis-001.json'));
  const sig = pkt.signature;
  delete pkt.signature;
  console.log(verify(canonicalize(pkt), sig, pkt.submittedBy));
"