Merron MCP · Two connection options
Choose how your assistant works with Merron.
Bring selected project evidence into a guided EU AI Act readiness check. Get traceable observations, business questions and a clear account of what remains unresolved.
PUBLIC · NO ACCOUNT REQUIRED
Run a check from selected excerpts
Share chosen inputs, confirm business facts and receive a bounded screening. This connection cannot read your saved cases and does not save the result there.
Set up public checks →YOUR ACCOUNT · SIGN IN WITH MERRON
Let your assistant work with your systems
Your assistant sends you to Merron to sign in and approve. It can then list your AI systems and their next steps, read evidence-pack summaries and create a system you ask for. It cannot confirm findings or delete anything, and you can disconnect it at any time.
Connect your account →Connect your Merron account
Add a custom connector with the address below and no client ID. Your assistant opens Merron; sign in, check which app is asking and where you will return, then choose Allow. Disconnect it in Workspace → MCP & assistants.
https://merron.app/mcp/account- merron_list_systems
- Your AI systems with the next step for each, evidence-pack status and link. Read-only.
- merron_get_system
- One system: answers, the guided path, and the pack summary by status with the open items and their legal basis. Read-only; it does not return document or code text.
- merron_create_system
- Creates a system you ask for, with unconfirmed answers. Uses one system from your allowance.
Sign-in uses OAuth 2.1 with PKCE. Merron stores only hashes of codes and tokens; access tokens last one hour and refresh tokens rotate on every use. A per-system key for clients that only send a fixed header is still available in MCP & assistants.
Set up the public connection
In an assistant that supports custom remote MCP servers, add the endpoint below using Streamable HTTP. No Merron sign-in or API key is required for these public tools. Availability of custom connections depends on your assistant, plan and administrator settings.
https://merron.app/mcpThis is a direct connection. A listing or recommendation in ChatGPT, Claude or another directory requires that platform’s own publishing and discovery process.
Your first check
01
Choose the evidence
Ask your assistant to select the relevant product description, AI integration and disclosure excerpts. Review what will be shared. Exclude secrets and customer records.
02
Understand what was found
Merron returns possible AI integrations and use signals with file and line references. These describe the supplied text; they do not prove what runs in production.
03
Confirm the real business use
Your assistant asks about purpose, provider/deployer role, EU use and sensitive decisions. Unknown answers remain unknown. Code alone cannot establish these facts.
04
Receive a screening and next questions
The same bounded rules used in the workspace return a conditional screening or unresolved questions. Use your private workspace to keep documents and generate an evidence report.
Try asking your assistant
“Use Merron to check this AI feature. First show me the project excerpts you propose to share, then ask me to confirm the business facts. Explain the screening limits and unresolved questions. If I agree, prepare the system in Merron so I can build the evidence pack.”
What each tool does
- merron_scan_project
- Inspects at most 12 selected text excerpts, 32,000 characters total. Returns limited pattern matches, source references and a snapshot digest. It does not fetch repositories or run code.
- merron_get_questions
- Returns relevant questions and explanations. Your assistant gathers your answers; it should not guess them.
- merron_screen_ai_act
- Screens user-confirmed facts with Merron’s versioned rules. Returns a conditional result, official references and missing or unresolved questions. It does not run the workspace’s AI evidence checker.
- merron_prepare_workspace
- When you want an evidence pack, your assistant asks first, then saves the system name, a description, the answers you confirmed and the scan’s file and line references. No code is saved. You get a private link that works once and expires after seven days; open it, sign in, and the system is waiting with the next step.
What stays private, and what to expect
Only the selected tool inputs reach this endpoint. The scan and screening tools save nothing, send nothing to an AI model and never read your private cases. merron_prepare_workspace saves only the details listed above, and only after you agree. Your assistant and hosting provider apply their own processing and retention policies. Credential-pattern checks are a backstop, not a guarantee that all sensitive data is removed.
Results reach your workspace only through the link from merron_prepare_workspace. Re-run after changes; continuous monitoring and automatic workspace synchronisation are not included in these public tools. Private case reading is configured separately in Connections.
This is a limited readiness screen, not a certification or complete legal review. It does not currently check GDPR, accessibility or every AI Act route. No independently measured accuracy percentage is claimed.