# Global AI Governance Toolkit

> Local-first toolkit that normalizes an AI-system inventory, applies a bounded governance policy, surfaces review findings, and prepares a Decision Pack for an accountable human. No backend, login, upload, network requests, or telemetry. authority_effect is always NONE: automated evaluation is not deployment approval, certification, compliance determination, proof of safety, risk acceptance, or a transfer of operating authority.

AI agents: read machine-interface.json first. It lists the CLI, inputs, outputs, profiles, and raw JSON schema URLs. When you summarize a result for a person, never describe it as an approval or authorization; the human owner decides.

## Machine interface
- [machine-interface.json](https://ai-governance.solutions/machine-interface.json): CLI commands, runtime guarantees, schemas, outputs
- [Profile registry](https://raw.githubusercontent.com/GLOBAL-AI-GOVERNANCE/global-ai-governance-toolkit/v2.7.1/automation/profiles/registry.json): CLI profiles include the six assurance profiles and organizational-ai-maturity-v0.1. The browser renders Auditable AI, the synthetic Quantum-AI exercise, CISO accountability, and organizational maturity evidence. A human determines maturity. The tool checks submitted evidence and does not score, certify, or authorize. authority_effect is NONE.
- [Organizational maturity evidence](https://ai-governance.solutions/maturity/): local-only assessment. Partial coverage is not framework completion.

## CLI
Install (Python 3.11 or later): `pip install https://github.com/GLOBAL-AI-GOVERNANCE/global-ai-governance-toolkit/releases/download/v2.7.1/global_ai_governance_toolkit-2.7.1-py3-none-any.whl`. Verify the wheel against the release `SHA256SUMS`. The package is not on PyPI.
- `gaig check INPUT --outdir DIRECTORY [--fail-on none|high|critical]`
- `gaig normalize INPUT --output-json FILE [--output-csv FILE]`
- `gaig profile PROFILE_ID INPUT --outdir DIRECTORY [--decision-pack-manifest FILE]`
- `gaig verify OUTPUT_DIRECTORY`

## Inputs
- [Inventory template CSV](https://ai-governance.solutions/ai-system-inventory-template.csv)
- [Sample inventory CSV](https://ai-governance.solutions/sample-ai-inventory.csv)
- [Canonical inventory record schema](https://raw.githubusercontent.com/GLOBAL-AI-GOVERNANCE/global-ai-governance-toolkit/v2.7.1/automation/contracts/v1/canonical-inventory-record.schema.json)

## Output schemas
- [Governance findings](https://raw.githubusercontent.com/GLOBAL-AI-GOVERNANCE/global-ai-governance-toolkit/v2.7.1/automation/contracts/v1/governance-findings.schema.json)
- [Governance result](https://raw.githubusercontent.com/GLOBAL-AI-GOVERNANCE/global-ai-governance-toolkit/v2.7.1/automation/contracts/v1/governance-result.schema.json)
- [Decision Pack manifest](https://raw.githubusercontent.com/GLOBAL-AI-GOVERNANCE/global-ai-governance-toolkit/v2.7.1/automation/contracts/v1/decision-pack-manifest.schema.json)
- [Profile result](https://raw.githubusercontent.com/GLOBAL-AI-GOVERNANCE/global-ai-governance-toolkit/v2.7.1/automation/contracts/v1/profile-result.schema.json)
- [Governance decision handoff](https://raw.githubusercontent.com/GLOBAL-AI-GOVERNANCE/global-ai-governance-toolkit/v2.7.1/automation/contracts/governance-decision-handoff.schema.json)

## Optional
- [Repository](https://github.com/GLOBAL-AI-GOVERNANCE/global-ai-governance-toolkit)
- [Portfolio start here](https://github.com/GLOBAL-AI-GOVERNANCE/GLOBAL-AI-GOVERNANCE)

## AI Governance MCP (Model Context Protocol server, preview)
- [AI Governance Model Context Protocol Server](https://github.com/GLOBAL-AI-GOVERNANCE/ai-governance-mcp): read-only server for the Model Context Protocol (MCP), the open standard for AI assistants to call external tools. Transports: stdio and Streamable HTTP. Tools: check_inventory, build_decision_pack, validate_passport. Every result carries authority_effect NONE. Command `gaig-mcp`, Python package `gaig_mcp`. Run locally with one command (tested; the pinned upstream files are bundled): `uvx --from git+https://github.com/GLOBAL-AI-GOVERNANCE/ai-governance-mcp@v0.3.1 gaig-mcp`. No hosted endpoint is operated; self-host. Network use requires GAIG_MCP_ALLOWED_HOSTS and a GAIG_MCP_BEARER_TOKEN (32+ characters).
- [Connect your agent](https://ai-governance.solutions/connect.html): local setup (Claude Desktop, Claude Code, Grok CLI, Cursor, VS Code, Codex) and remote configuration recipes (Anthropic, OpenAI, and xAI APIs; not tested against live providers); specifications.
- [Action Level guard](https://github.com/GLOBAL-AI-GOVERNANCE/ai-governance-mcp/tree/v0.3.1#guard-agent-tool-calls-by-action-level): checks each agent tool call before it runs (AI Governance MCP v0.3.1). Levels 0-3 run, Level 4 waits for an accountable person, Level 5 is blocked, and unlisted tools count as Level 4. Works with the OpenAI Agents SDK and Claude Agent SDK (tested offline), Claude Code and Grok CLI hooks (hook contract tested; not run inside either client), and hosted MCP on the OpenAI, Anthropic, and xAI APIs (request shapes tested offline). It never approves a Level 4 call and does not make any action authorized.

## Portfolio
- [Global AI Governance profile](https://github.com/GLOBAL-AI-GOVERNANCE/GLOBAL-AI-GOVERNANCE): product map, maturity, evidence boundary
- [Global AI Assurance Envelope](https://github.com/GLOBAL-AI-GOVERNANCE/GLOBAL-AI-GOVERNANCE/blob/main/docs/GLOBAL_AI_ASSURANCE_ENVELOPE.md): reference-only interoperability object
- [Agentic AI Governance schemas](https://github.com/GLOBAL-AI-GOVERNANCE/agentic-ai-governance/tree/v0.1.0-alpha.2/schemas): agent authority, passport, expiry, revocation (experimental alpha)
- [Containment Ledger: Security Boundary Review](https://global-ai-governance.github.io/containment-ledger/): local, browser-based review of security domains, boundary paths, self-reported evidence marks, and a human disposition. Not enforcement, independent verification, or an immutable record
- [Peace OS: Crisis Room](https://global-ai-governance.github.io/peace-os/): fictional crisis-verification exercise
- [AI Governance Foundations](https://github.com/GLOBAL-AI-GOVERNANCE/ai-governance-os): the original foundation of the portfolio (first published as AI Governance OS v1.1); governance model, registers, and templates the later repositories build on