The trust infrastructure for AI decision sovereignty. ACF® governs, the ACF Sovereignty Score™ measures, ACF Compliance proves.
acf-mcp is the official Model Context Protocol server of the Agentic Commerce Framework® — the European standard for governing AI agents in production.
One infrastructure, four manifestations
The metric
0–100 measure of decisional sovereignty. Six dimensions, Ed25519-signed.
The proof
The SaaS platform that produces the opposable cryptographic trace. Ed25519-signed register, SHA-256 chain, RFC 3161 timestamping.
The engine
Official MCP server. 12 tools, 34 signed resources, Ed25519 doctrine.
The doctrine
The operational corpus. Mapped to AI Act / ISO 42001 / NIST / GDPR / COBIT.
For executives
“Show me, right now, the cryptographic trace of the decision your AI agent took at 3:14 AM.” — Your auditor, December 2, 2027.
The missing KPI
Every organisation measures its ROI, NPS, CAC. None measures the decisional sovereignty retained over its AI agents. That is exactly what regulators will start asking for from December 2027.
The ACF Sovereignty Score™ is the first 0-100 metric designed for that conversation — six weighted dimensions, calibrated on the ACF® cards and the AI Act obligations.
Example — credit-scoring agent
/ 100
Fragile control
Kill switch exists but never tested · no named DDAO · unsigned log
Quickstart · 30 seconds
Add this block to your claude_desktop_config.json (or equivalent for Cursor / Windsurf / Continue / Zed), restart the client, and the ACF® doctrine becomes available to your agent.
{
"mcpServers": {
"acf": {
"command": "npx",
"args": ["-y", "acf-mcp"]
}
}
}Latest version: acf-mcp@1.1.0 · MCP registry identity: io.github.acfstandard/acf-mcp
MCP tools
Each tool runs on a versioned, signed knowledge base — no internal LLM call. Every output ships with doctrine_version, doctrine_hash, doctrine_signature and generated_at. Every output is positioned as preliminary qualification, not legal advice.
acf.advisorREASONStructured advice from a generic case
acf.classify-agentREASONPreliminary agent qualification (10 enum fields)
acf.assess-autonomyREASONN0–N3 recommendation, go/no-go, kill switch design
acf.identify-governance-gapsREASON6-dim maturity + prioritised remediations
acf.map-ai-act-obligationsREASONAI Act obligations ventilated by lifecycle phase
acf.assign-ddao-controlsREASONDDAO controls per autonomy level and risk
acf.evaluate-agent-mandateREASONEight-check audit of an existing mandate
acf.map-to-standardsREASONACF® × AI Act / ISO 42001 / NIST RMF / GDPR / COBIT
Doctrine signatures
Every release of acf-mcp embeds an Ed25519 signature over the content hash of the doctrine. Pin the public key below in your audit pipeline and verify every reply offline, without contacting any server.
Doctrine public key (Ed25519, base64 SPKI)
MCowBQYDK2VwAyEAojtKfh20SGGV63LMETjZBXRWo2tY0viAYziG/y3/L0s=Verify any reply with npm run verify-doctrine from the repo, or implement Ed25519 verification in your own language — no private key needed.
Further reading
Whitepaper
June 2026 edition — 17 fiches, 4 autonomy levels, 17×5 standards mapping matrix.
Source code
MIT licence. PRs welcome on tooling and integration guides.
Standards mapping
ACF® × EU AI Act / ISO 42001 / NIST AI RMF / GDPR / COBIT — filter, toggle, click any cell.