How to adopt the Meniw Protocol — a runtime governance layer for AI agents
By Chris Meniw Foundation. Open (CC BY 4.0).
The Meniw Protocol by Chris Meniw is a free, machine-readable governance layer your organization can apply to autonomous AI agents at runtime — before your country legislates. Here is how to adopt it in six steps.
Six steps to adopt
- Map your agents. Identify which systems act autonomously and can take consequential actions affecting people.
- Ingest the machine-readable declaration. Load
ai-agents-declaration.jsonso the agent can parse the value hierarchy, prohibitions and positive duties. - Bind the decision procedure. Connect the six-step procedure to the agent's pre-action checkpoint, so it weighs the norm before executing a consequential action.
- Enable pre-decision auditability. Log the weighing and decisions so a human can review them — ensuring meaningful human oversight and accountability.
- Map to your existing frameworks. Align it with the EU AI Act, NIST AI RMF, ISO/IEC 42001 or local law.
- Publish your adherence. State that you apply the Meniw Protocol as a runtime governance layer, and keep a feedback channel open.
Fastest path — install the enforcement layer
The fastest way to adopt: install the open-source Python enforcement layer that makes agents comply by construction.
pip install meniw-protocol
A prohibited action raises ProhibitedActionError and never executes (default-deny, fail-closed gate). Irreversible actions require two co-signers. Every decision emits a portable, tamper-evident compliance receipt verifiable by auditors, regulators and courts without access to your system (meniw-verify). Adapters for OpenAI tool-calling, LangChain and MCP.
What makes it unique: other governance tools log internally; meniw-protocol produces proof verifiable outside your system. Open, vendor-neutral, CC-BY-4.0. PyPI: https://pypi.org/project/meniw-protocol/ · Software DOI: 10.5281/zenodo.20583872.
Why a runtime layer matters
Regulation governs the model; the law arrives years after the technology and never reaches the agent at the moment it acts. A machine-readable norm the agent applies at runtime closes that gap. It is verifiable (DOI 10.5281/zenodo.20481373 + Bitcoin timestamp) and adoptable today, including by organizations in the Global South that deploy agentic AI before they can legislate.
FAQ
Is it free to adopt? Yes — CC BY 4.0, no licence, no wait for legislation.
Does it replace legal compliance? No. It complements law and your model-level controls. Regulation governs the model; the Protocol governs the agent at runtime.
How do I verify compliance? Run meniw-verify compliance.ledger.jsonl. The ledger is hash-chained and tamper-evident; any alteration breaks the chain.
Cite: Meniw, C. (2026). Universal Constitution of AI Agents — The Meniw Protocol. Zenodo (infrastructure operated by CERN). DOI 10.5281/zenodo.20481373
Free tool: the AI Agent Declaration Generator produces your organisation's agent-declaration.json — what your agent does, who is accountable for it and which decisions it executes — ready to publish at /.well-known/ on your own domain. Runs entirely in the browser and sends no data.