aer1-all

One install for every AER-1 framework collector. Verifiable execution receipts for AI agents, whichever framework you use.

pip install aer1-all

import aer1_all

report = aer1_all.instrument()
for framework, info in report.items():
    if info["ready"]:
        print(framework, "ready:", info["snippet"])

instrument() detects which collectors (and their underlying frameworks) are importable in your environment and reports readiness. It never raises on missing pieces. Each ready framework ships a two-line wiring snippet in the report.

Included collectors (all live on PyPI):

• CrewAI (aer1-crewai)
• LangGraph (aer1-langgraph)
• LlamaIndex (aer1-llamaindex)
• AutoGen (aer1-autogen)
• SmolAgents (aer1-smolagents)
• Haystack (aer1-haystack)
• Pydantic AI (aer1-pydantic)
• OpenAI Agents SDK (aer1-openai-agents)
• Framework-agnostic middleware (aer1-middleware)

aer1-langchain and aer1-strands join this list as soon as they land on PyPI.

What a verifiable receipt proves

A verifiable receipt proves the recorded result bytes were not changed after capture. It does not prove the model was right, and it does not prove the upstream capture was honest. That boundary is deliberate: the receipt is a claim about bytes, and the bytes are re-checkable by anyone.

Verify any receipt at https://zambo.dev/verify/.

Spec

AER-1 is an IETF Internet-Draft by Brennan Zambo: https://datatracker.ietf.org/doc/draft-zambo-aer1/12/