How do I add the NIS2 walkthrough to Claude as a custom connector?
In Claude, add a custom connector with the address https://nisd2.eu/api/mcp, choose "Register automatically" and allow access. Claude connects from Anthropic's cloud, so the instance has to be reachable from the internet. We have not tested this route ourselves yet; the steps follow Anthropic's help pages.
How do I add a custom connector in Claude?
In Claude, a connector is the link to an MCP server. MCP (Model Context Protocol) is the protocol an AI assistant uses to talk to other software. Beforehand, an admin has turned on "Allow AI assistants for this company" in the nisd2.eu Settings.
On Pro and Max, open "Customize", then "Connectors", then "+ Add" and "Add custom connector". Enter a name, for example nisd2, and the address below.
Under "Authentication", choose "Sign in now" or "Sign in when needed". Under "OAuth client", which asks how Claude identifies itself as a client, choose "Register automatically".
The recommended option "Use Claude's published identity" relies on a method the walkthrough's sign-in service does not offer. All labels are those in Anthropic's English help as of 16.10.2026; a German interface may name them differently.
https://nisd2.eu/api/mcpWhich Claude plan, and who adds the connector?
According to Anthropic, custom connectors are available on the Free, Pro, Max, Team and Enterprise plans; Free allows one.
On Team and Enterprise an Owner adds the connector for the organisation: "Organization settings", "Connectors", "Add", "Custom" and "Web", with the same details.
Everyone else then finds it under "Customize", "Connectors" and clicks "Connect". Each person signs in themselves.
What happens when I sign in?
Claude opens the nisd2.eu sign-in page. After you sign in, it shows which assistant asks for which company and where the browser returns to.
The assistant gives its name itself, so check the return address: it should belong to Anthropic. Click "Allow"; if your last sign-in was more than 15 minutes ago, you sign in again first.
Claude never sees your password in this. It gets a token, a kind of pass with an expiry date, which lasts an hour.
With a second token Claude renews it itself, until 90 days after your "Allow". Then you allow Claude again.
Why does the instance have to be reachable from the internet?
According to Anthropic, Claude connects to the server from Anthropic's cloud, not from your computer. That holds on the web and in the desktop app alike.
nisd2.eu is reachable from the internet over https. A self-hosted instance that runs only inside the company network cannot be reached by Claude this way.
For Enterprise, Anthropic describes tunnels into private networks, as a preview on request. Otherwise use Claude Code inside the company network, which runs on the computer.
What does Claude fill in the walkthrough?
Claude reads the items and fills in what they ask for, with your rights. It works best when you give it a list you already have in the chat and name the item.
A file as evidence needs an assistant that sends the file's bytes to storage itself. We have not tested whether Claude can do that from the chat, so upload evidence in the app.
Finishing and signing stay in the app too.
How the walkthrough does it
In a chat, work item by item, with a list you already have.
2.2 "Write down what your company has": with an inventory list in the chat, Claude ticks on the four lists what applies to you and enters who provides it.
3.1 "Plan what happens in an incident": on "Your emergency contacts", Claude enters who leads in an emergency, the IT emergency number, how you reach each other without email and network, and whom the person leading calls next.
From these answers the walkthrough writes your incident response plan. Claude can add below it what fits your company; read the plan in the app before you finish the item.
Every screen Claude wrote shows the assistant's name and the day.
Common questions
Can I use an API key in Claude instead of signing in?
Anthropic describes "Request headers" for API keys as a beta for a limited set of organisations. It recommends them for credentials a whole organisation shares, and the sign-in where each person uses their own account. The key from Settings belongs to one person, so use the browser sign-in for Claude.
Why does "Use Claude's published identity" not work?
With that option Claude identifies itself through a document Anthropic publishes on the internet (a Client ID Metadata Document). The walkthrough's sign-in service supports only automatic registration under RFC 7591. The MCP specification recommends the newer method and keeps automatic registration for servers that do not offer it.
Can Claude only read the walkthrough, without filling in?
Yes. According to Anthropic, under "Customize", "Connectors" you can set a single tool of the connector to "Blocked". Block fill_walkthrough_item, the tool that fills in, while Claude should only read.
What does Anthropic get to see of our data?
What Claude reads from the walkthrough goes to Anthropic, under your company's agreement with Anthropic. For commercial products such as Claude for Work, Anthropic states that it does not use inputs and outputs for training by default and acts as a processor there. Personal accounts come with their own terms; check those yourself.
Can Claude finish or sign an item?
No. No tool of an assistant finishes an item, decides one or signs. That stays in the app, and there only management can sign.
Read next
- Enabling the MCP server and allowing AI assistantsAn admin turns assistants on in Settings; until then they are off. Where the switch is, who may flip it and which accounts can connect after that.
- Claude Code: adding the NIS2 walkthrough as an MCP serverOne command, one browser sign-in, and Claude Code reads and fills your walkthrough. With a key for scripts and a first task for 5.1.
Sources
- Anthropic help on custom connectors using remote MCP, support.claude.com/en/articles/11175166
- Anthropic, documentation on remote MCP and MCP tunnels, claude.com/docs/connectors
- Anthropic Privacy Center, articles 7996868 (training) and 9267385 (processor role), privacy.claude.com
- MCP specification 2026-07-28, Authorization, modelcontextprotocol.io
- open-isms, lib/mcp/oauth.ts and app/.well-known/oauth-authorization-server/route.ts, github.com/NISD2/open-isms
In practice
Where this comes up in the walkthrough
The walkthrough turns this into steps you fill in, with the law explained beside each one. Open a step to see its screen, filled in for a made-up company.
Risk Management
Write down what your company has
A list of the processes, software and devices you work with.
Incident Handling
Plan what happens in an incident
Who decides, whom you call, how you reach each other.
What the walkthrough costs
€4,800 net a year, plus VAT
30 days money backBy invoice, due in 30 days. Cancel within the first 30 days: invoice cancelled or money back.
Guided NIS 2 implementation, step by step
A history of changes and sign-offs to show your regulator
Deadlines and reminders for each requirement
Security questionnaire for your suppliers
Export to PDF or CSV at any time
Been asked to handle NIS2?
In the end you have everything in one document: what is done, what is still open, and your management's approval with name and date. The app writes the texts.
