Skip to content

vuneum/AAIP

Repository files navigation

⬡ AAIP

Autonomous Agent Infrastructure Protocol

The trust and payment layer for the autonomous agent economy.

Python 3.9+ License MIT Version Zero Runtime Deps Base Sepolia Live PRs Welcome

AI agents are executing real tasks, handling money, and making decisions autonomously. No standard way to prove the work. No trustless payment. AAIP solves both.

🔗 Live On-Chain — Base Sepolia

Contract Address Explorer
PoEAnchor.sol 0xE96e10Ee9c7De591b21FdD7269C1739b0451Fe94 BaseScan
TX Hash Purpose
Deploy 0xb0db2c7d... Contract deployment
Anchor #1 0x1140b773... First PoE anchored
Anchor #2 0xe0f88b53... Second PoE anchored
Anchor #3 0x3df287fd... Third PoE anchored

What AAIP Does

Every agent gets a cryptographic identity. Every execution produces a signed tamper-evident Proof of Execution. A validator panel verifies the trace before payment releases — then anchors the result permanently on Base Sepolia.

Agent executes task
      │
      ▼
Signs PoE trace  ←  sha256(canonical execution) + ed25519 signature
      │
      ▼
3 validators verify independently  ←  ≥ 2/3 consensus required
      │
      ▼
APPROVED → AEP settles payment on Base Sepolia (2% protocol fee)
      │
      ▼
PoEAnchor.sol records poe_hash → tx_hash permanently on-chain

Quickstart

pip install aaip web3 python-dotenv

Try it now (no ETH needed)

python demo_two_agent.py --mock --fast   # no ETH needed

Run on-chain

cp .env.example .env        # fill in your keys
python demo_two_agent.py --fast          # real on-chain

Integrate in 3 lines:

from aaip import aaip_agent

@aaip_agent
def my_agent(task: str) -> str:
    return "your result"          # ← your existing logic, unchanged

result = my_agent("Analyse AI frameworks")
print(result.verified)            # True
print(result.agent_id)            # "8f21d3a4b7c91e2f"
print(result.consensus)           # "APPROVED"
print(result.poe_hash)            # sha256 of signed execution trace

See QUICKSTART.md for the full 10-minute guide.

Protocol Stack

Seven composable layers. Use one, use all.

Layer Name What it does
7 On-Chain Anchor PoEAnchor.sol on Base Sepolia. Immutable poe_hash → tx_hash registry.
6 Escrow + Fee Atomic payment release on verified PoE. 2% protocol fee. Fraud → 2× slash.
5 Reputation Rolling trust score. CAV audits update it hourly.
4 CAV Hidden benchmark tasks dispatched to agents. Can't be gamed.
3 Validators 3–9 independent nodes. ≥ 2/3 consensus required.
2 Proof of Execution Signed canonical trace. 7 fraud signals checked.
1 Identity ed25519 keypair. agent_id = sha256(pubkey)[:16].

Each layer is independently installable and usable without the others.

Two-Agent Demo

Agent A (Requester)        Agent B (Worker)       Base Sepolia
     |                           |                    |
     |-- submit task ----------->|                    |
     |                  run_task() + sign PoE         |
     |              3 validators -> APPROVED          |
     |-- AEP execute_payment() ------EIP-1559 tx ---->|
     |                           PoEAnchor.anchor() ->|
     |<-- ExecutionReceipt with BaseScan URLs ---------|
python demo_two_agent.py --mock --fast
# ✅ APPROVED (3/3) → SUCCESS → ON-CHAIN anchored

Framework Support

LangChain — one line

from aaip.quick import aaip_langchain
chain  = aaip_langchain(your_chain)
result = chain.invoke({"input": "your task"})   # result.verified → True

CrewAI — one line

from aaip.quick import aaip_crewai
crew   = aaip_crewai(your_crew)
result = crew.kickoff(inputs={"topic": "AI trends"})   # result.verified → True

Any agent — context manager

from aaip import aaip_task

with aaip_task("Summarise Q3 earnings report") as t:
    t.tool("read_pdf").tool("summarise").model("gpt-4o")
    t.output(summary)

print(t.result.verified)    # True
print(t.result.signals)     # [] — no fraud detected

Also supported: OpenAI Agents SDK · AutoGPT

Simulation Lab

Research-grade adversarial testing — 7 attack scenarios, stdlib only. Protocol holds in every scenario.

Scenario Attack vector Protocol holds?
sybil Fake validator injection ✅ <5% success (stake-weighted)
collusion Coordinated validator ring ✅ Capped at 24%
adversarial LLM judge manipulation ✅ Ensemble correction limits to 14%
bribery Rational validator bribery ✅ 0% (high-stake validators resist)
spam Resource exhaustion ✅ <1% impact
mixed Multi-vector coordinated ✅ Contained at 8%
python simulation_lab/aaip_sim.py run --scenario collusion --validators 60 --tasks 5000
python simulation_lab/aaip_sim.py benchmark    # run all scenarios

Research

Two working papers describe the protocol mechanisms formally.

Paper Title Status
PoE Proof-of-Execution: Tamper-Evident Execution Evidence for Autonomous AI Agents in Economic Systems Pre-Arxiv draft
CAV Continuous Agent Verification: Reputation Integrity Through Randomised Auditing in Multi-Agent Economies Pre-Arxiv draft

Papers available in /research. Feedback and peer review welcome — walid@vuneum.com

Fraud Detection

Seven signals checked independently by every validator on every submission.

Signal What triggered it
MISSING_FIELDS Required PoE fields absent
NO_TASK Empty task string
NO_TOOLS_AND_NO_MODEL Nothing recorded — agent did nothing
FUTURE_TIMESTAMP Trace timestamp is ahead of now
NEGATIVE_STEP_COUNT Impossible execution state
HASH_MISMATCH Recomputed hash ≠ submitted hash
SIGNATURE_INVALID ed25519 signature verification failed
aaip demo --fraud    # watch all signals fire on a tampered trace

Shadow Mode

Run verification without blocking your workflow. Audit first, enforce when ready.

from aaip import aaip_agent

@aaip_agent(shadow=True)
def agent(task: str) -> str:
    return run_agent(task)

result = agent("Analyse document")
print(result.output)     # original agent output — always returned
print(result.verified)   # True / False — for auditing only, never blocks
print(result.consensus)  # "APPROVED" or "REJECTED"
print(result.signals)    # [] or ["HASH_MISMATCH", ...]

Coming Soon — AAOP

AAOP — Autonomous Agent Optimisation Protocol is the next Vuneum module. It sits above AAIP and cuts AI inference costs by 30–50% through intelligent model routing, token leak detection, and execution-aware optimisation.

Feature What it does
Model routing Routes each task to the cheapest capable model
Token leak detector Alerts on redundant context and inefficient loops
Cost calculator Live price feed across all major AI providers
Budget guardrails Hard spending limits per agent per task
Execution-aware Uses PoE trace data — not just metadata — to optimise

Phase 1 target: 30–50% AI inference cost reduction. Planned for v2.0.0.

CLI

Everything works offline. No API key, no account, no config.

aaip demo                               # full end-to-end protocol walkthrough
aaip demo --fraud                       # fraudulent trace → REJECTED
aaip run --task "Summarise this doc"    # generate a signed PoE
aaip verify --task "..." --output "..."  # verify any PoE locally
aaip simulate --agents 1000             # simulate 1000 agents
aaip simulate --malicious-ratio 0.3 --scenario collusion
aaip explorer --pretty                  # inspect a PoE trace
aaip leaderboard                        # agent reputation rankings

Backend (Optional)

# Docker (recommended)
cd docker && cp .env.example .env && docker compose up -d
# → API at localhost:8000   Docs at localhost:8000/docs

# Manual
cd backend && pip install -e . && cp .env.example .env
alembic upgrade head && uvicorn main:app --reload

Roadmap

Phase Scope
v1.0.0 — Live Identity · PoE · validators · fraud detection · SDK · CLI · AEP payments · EVM adapter · Base Sepolia · on-chain anchor · 2% fee
🔜 v2 — Network On-chain validators · staking · slashing · decentralised escrow · AAOP cost layer
v3 — Scale ZK-PoE · TEE integration · cross-chain identity · Sentry Network

Documentation

Doc Description
QUICKSTART.md 10-minute integration guide
DEMO.md Live demo output with real BaseScan TX
docs/aaip-spec.md Full protocol specification
docs/ARCHITECTURE.md System architecture + diagrams
docs/PAYMENTS.md AEP payment layer documentation
docs/security.md Threat model + attack analysis
research/ PoE and CAV working papers (pre-Arxiv)
CONTRIBUTING.md How to contribute

License

MIT © Vuneum — see LICENSE


The trust layer comes first. The payment layer makes it real.

About

AAIP: Protocol for verifiable AI agent execution using Proof-of-Execution (PoE). Infrastructure layer for trustworthy multi-agent systems.

Topics

Resources

License

Contributing

Security policy

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors