Global AI Governance
Govern AI before it acts.
Open-source governance for agent fleets, AI agencies, and AI-run companies, wherever they operate. Find the missing owner, monitoring, shutdown path, or evidence before anything runs, and give the person accountable a Decision Pack and the final decision. Framework-neutral and jurisdiction-neutral.
- Open and local-firstMIT and Apache-2.0. The browser Toolkit reads your files on this device, with no upload or account.
- Any MCP agentA standard Model Context Protocol server, with setup for agents and APIs from OpenAI, Anthropic, and xAI.
- Mapped to frameworksCrosswalks to NIST AI RMF 1.0, ISO/IEC 42001, the EU AI Act, and frontier safety frameworks.
- Attested releasesToolkit
v2.8.3, MCPv0.5.1, Agentic AI Governancev0.1.0-alpha.6, Verified Vulnerability Governancev0.4.0, and Containment Ledgerv1.0.0ship SHA-256 checksums and build provenance.
Start
Start where you are
For anyone who uses, builds, deploys, oversees, researches, or regulates AI. A few examples follow; every step is open, so begin anywhere.
- You use AI appsChatGPT, Claude, Grok, Gemini, or Copilot. Check the AI you use and give it a ground rule, in your browser.
- You build AI organizations or agent fleetsInventory, passports, guards, fleet controls, shutdown tests, and vulnerabilities closed with evidence.
- You work at a frontier labCrosswalks, contracts, attested releases, and paths for independent review.
- Anyone elseTeams, enterprises, researchers, educators, public-interest bodies, and regulators: every piece in one map.
Or go by how you work: in this browser, from the command line, or inside your agent.
For AI agents: start at AGENT_SETUP.md, JOURNEYS.md, or llms.txt. Every command comes with the result to expect.
The suite
From inventory to verified closure
Toolkit
Turn an AI-system inventory into findings and a Decision Pack for the person accountable.
v2.8.3 · stableAI Governance MCP
The same checks, callable by your agents over Model Context Protocol (MCP). Read-only.
v0.5.1 · releasedAgentic AI Governance
Agent Trust Passports and Action Levels: who an agent acts for, what it and its sub-agents may do alone, and when that ends.
v0.1.0-alpha.6 · availableVerified Vulnerability Governance
Track each vulnerability from owner to fix to retest, and close it only with evidence reviewed by a separate role.
v0.4.0 · releasedAI Cyber Resilience Framework
Limit the blast radius of AI-accelerated cyber risk, protect recovery, and keep critical work running.
v0.2.0 · releasedContainment Ledger
Review a system's security boundaries in your browser: domains, paths, evidence, and a human disposition.
v1.0.0 · releasedPeace OS: Crisis Room
A crisis-room simulator for weighing evidence, protecting civilians, and keeping the final decision with people. Scenarios are synthetic.
v0.3.0 · releasedGoverned Systems Administration
Contracts for human-governed, AI-assisted systems administration, specified ahead of any execution.
specification · no executionAI Governance Foundations
The governance model, registers, and templates the rest of the suite builds on.
v1.1.1 · stable
Toolkit · v2.8.3
Run a check in this browser.
List each AI system or agent in a simple inventory. The Toolkit shows which ones lack a named owner, monitoring, a shutdown path, or evidence, and prepares a Decision Pack so the person accountable can make and record the decision. No install, no upload, no login.
Advisory vs enforcement. Browser and CLI results are advisory. Runtime enforcement is the MCP guard and hooks.
What goes inAn AI-system inventory
What comes outGovernance gaps and a Decision Pack
WhyTo support an accountable human decision
Run an example → Download and fill the template → Load the CSV → Review what needs attention → Download the Decision Pack
Assess organizational AI maturity evidence
Both examples are synthetic inventories. The example with gaps includes an autonomous agent with no named owner, monitoring, or shutdown path.
Machine interface · Versioned contracts and agent entry point