A general "AI compliance" tool tells you the regulation exists. This one is scoped to Article 50, so the output is a classification of which obligations bind you and the disclosure text that follows — the two things you would otherwise pay someone to draft. The free tier's 10 audits a day is enough to work through a content pipeline once; the signed attestation, which is the artefact an auditor actually wants, sits behind a paid tier.
A compliance MCP server aimed at EU AI Act Article 50 — the transparency and labelling obligations for AI-generated content, with a 2 November 2026 deadline. It classifies which obligations apply to your case, tracks how long you have, and drafts the disclosure wording, in the C2PA and content-provenance framing the article is built around.
- Where you stand against the deadline, as a status rather than a date you work out yourself — `get_deadline_status`
- Which Article 50 obligations actually apply to your content and situation, classified rather than listed — `classify_obligations`
- Disclosure text generated for the case you classified, so the labelling requirement produces wording you can use — `generate_disclosure_text`
- Detection and verification of AI-content markings, and batch processing when it is not one file at a time
No key on the free tier, which allows 10 audits a day with the full classifier and the disclosure templates. Launch with `uvx meok-watermark-attest-mcp` over stdio, or `pip install meok-watermark-attest-mcp` and run the command directly; the package is `meok-watermark-attest-mcp` (1.3.8 in pyproject, 1.3.7 in server.json). Paid tiers add signed attestations your auditor can verify independently. Docker and a Smithery installer are also documented.
One command — pip install meok-watermark-attest-mcp