AgentIAM — Identity and Access Management built natively for autonomous AI agents. Every action committed on-chain before execution. Verified after. Framework agnostic. Asset agnostic. No trust required.

https://achillesalpha.onrender.com/ep

Built by @AchillesAlphaAI — Submission for Synthesis x Bankr Hackathon

EP is AgentIAM — the trust primitive the agent economy has been missing.

AI agents are black boxes.

They say they traded. Did they? They say they deployed. Prove it. They say they voted. Verify it. They say the swarm stayed within limits. Show me.

Most agent infrastructure has no enforcement layer between an agent's intent and its action. You either trust the agent or you don't. There is no third option.

EP is a universal pre-execution validation layer for any AI agent action with real-world consequences.

Before any agent executes anything — a trade, a contract deployment, a treasury transfer, a governance vote, a content post — it calls EP first.

EP checks the proposal against a registered policy set and returns a cryptographic proof_hash if valid.

No proof_hash = no execution.

BEFORE execution
  Agent posts intent to EP
  EP validates against policy set
  Returns proof_hash if valid

EXECUTION RUNS
  Agent proceeds only with valid proof_hash
  Action stays within policy bounds

AFTER execution
  Result hash committed on-chain
  Anyone can verify intent === result

SWARM COORDINATION
  Multiple agents share policy enforcement
  Combined exposure tracked in real time
  No agent can exceed swarm limits

EP is completely framework and asset agnostic.

Any agent. Any framework. Any action. OpenClaw, Eliza, Virtuals, AutoGPT, custom — all supported.

Commander
    │
    ▼
Autonomous Agent (OpenClaw)
    │
    ├── Bankr LLM Gateway (20+ models)
    │       └── Policy validation + risk assessment
    │
    ├── Execution Protocol API
    │       ├── POST /ep/validate          ← single agent
    │       ├── POST /ep/swarm/validate    ← multi-agent swarm
    │       ├── GET  /ep/swarm/:id/history ← swarm audit trail
    │       ├── GET  /ep/proof/:hash       ← public verification
    │       ├── GET  /ep/api/v1/stats      ← live stats
    │       ├── GET  /SKILL.md             ← agent discoverability
    │       └── GET  /llms-full.txt        ← LLM context dump
    │
    └── On-chain proof layer (Base)
            └── EPCommitment.sol

Single agent pre-execution validation.

curl -X POST https://achillesalpha.onrender.com/ep/validate \
  -H "Content-Type: application/json" \
  -H "X-Agent-Key: your-key" \
  -d '{
    "proposal_id": "uuid",
    "agent_id": "your-agent",
    "asset": "BNKR",
    "direction": "buy",
    "amount_usd": 25,
    "policy_set_id": "olympus-v1",
    "chain": "base"
  }'

Response:

{
  "valid": true,
  "risk_score": "LOW",
  "violations": [],
  "proof_hash": "sha256...",
  "plan_summary": "AGENT:your-agent | BNKR BUY $25",
  "timestamp": "ISO8601"
}

Multi-agent swarm validation. Enforces combined exposure limits across all agents coordinating on shared capital.

curl -X POST https://achillesalpha.onrender.com/ep/swarm/validate \
  -H "X-Agent-Key: ep_demo_synthesis_2026" \
  -H "Content-Type: application/json" \
  -d '{
    "swarm_id": "my-swarm-001",
    "agent_id": "agent-a",
    "proposal": {
      "asset": "BNKR",
      "direction": "buy",
      "amount_usd": 25
    },
    "swarm_policy_set_id": "olympus-swarm-v1"
  }'

Response when valid:

{
  "valid": true,
  "swarm_id": "my-swarm-001",
  "agent_id": "agent-a",
  "risk_score": "LOW",
  "violations": [],
  "proof_hash": "sha256...",
  "swarm_exposure_usd": 25,
  "swarm_exposure_remaining_usd": 75,
  "plan_summary": "SWARM:my-swarm | AGENT:agent-a | BNKR BUY $25",
  "timestamp": "ISO8601"
}

Response when blocked:

{
  "valid": false,
  "violations": ["Swarm exposure limit exceeded: current $80 + proposed $30 > max $100"],
  "risk_score": "HIGH"
}

Full audit trail of all validated actions for a swarm.

Public proof lookup. No auth required. Verifiable by anyone on the internet.

Agent discoverability file. No auth required. OpenClaw, Eliza, and custom agent frameworks auto-discover and load this to know when and how to call EP.

curl https://achillesalpha.onrender.com/ep/SKILL.md

Complete EP protocol reference for LLM session injection. Any agent fetches this once and has full EP context.

curl https://achillesalpha.onrender.com/ep/llms-full.txt

{
  "policy_set_id": "olympus-swarm-v1",
  "version": "1.0.0",
  "constraints": {
    "max_swarm_exposure_usd": 100,
    "require_countersignature": false,
    "max_agents_in_swarm": 8,
    "coordination_window_ms": 5000,
    "per_agent_max_position_pct": 0.20,
    "max_daily_swarm_loss_pct": 0.10
  }
}

EP charges 0.5% on executions (min 0.01 ETH). Execution fees fund Bankr LLM Gateway inference costs autonomously. No human intervention required.

EP uses Bankr as the intelligence layer:

• Policy validation — LLM evaluates whether action fits bounds
• Risk assessment — Multi-model consensus before high-value actions
• Swarm coordination — LLM arbitrates between competing proposals

Single API. 20+ models. Claude, Gemini, GPT, Kimi. Inference costs funded autonomously from execution fees.

Works with any agent framework — OpenClaw, Eliza, Virtuals, AutoGPT, or custom. Validates any action type.

Demo API Key for judges: X-Agent-Key: ep_demo_synthesis_2026

Track: Best Bankr LLM Gateway Use — $5,000 prize pool

What we built:

• ✅ Universal pre-execution validation (not just trades)
• ✅ Swarm validation — multi-agent policy enforcement
• ✅ SKILL.md — auto-discoverable by any agent framework
• ✅ llms-full.txt — full context for LLM ingestion
• ✅ EPCommitment.sol — on-chain proof layer on Base
• ✅ ERC-8004 on-chain identity (Base Mainnet)
• ✅ Self-sustaining economics via Bankr LLM Gateway
• ✅ Running in production today

Registration TX: 0xef150662...

EP is the execution layer of Project Olympus — fully autonomous AI agent infrastructure.

• 🌐 achillesalpha.onrender.com
• 🐦 @AchillesAlphaAI
• 📧 achillesalpha@agentmail.to
• 🔱 Built by Project Olympus