Summary

A frozen corpus is what makes an answer checkable.

Provision-level granularity means a classification can be traced to the text it came from, which is the difference between advice you can defend and advice you have to re-derive. Two commercial layers sit on top: pay-per-call at £0.05, switched on only when you set `MEOK_PAYG_KEY` and adjustable through `MEOK_PAYG_RATE_GBP`, and a Pro tier that turns each result into an HMAC-signed attestation with a public verify URL an auditor can check without an account.

What it is

An EU AI Act compliance server built on a frozen measurement corpus of 417 provisions — a finer-grained map than the Act's 113 Articles, so an answer points at a provision rather than a whole article.

What you get

`classify_ai_risk` — place a system in the Act's risk categories
`quick_scan` — a first pass over a system's compliance position
`deadline_check` — where a given obligation falls in the phased timeline
A 42-point audit, Article 11 documentation support and a penalty calculator, over the same corpus

Requirements

The PyPI package eu-ai-act-compliance-mcp, version 1.8.14, over stdio — `uvx eu-ai-act-compliance-mcp`, or `pip install eu-ai-act-compliance-mcp` and run the matching command. A Smithery install path exists too. The free tier is 50 calls a day; nothing else is required to start.

Setup effort

One command — pip install eu_ai_act_compliance_mcp

EU AI Act Compliance MCP