Skip to content
regensbot
Connect

Connect your assistant.

Regensbot is a remote MCP server at one URL. Clients with MCP sign-in need only that URL. Every other client sends an API key. Then your assistant audits URLs, reads your sites' checklists and cites the source of every finding.

https://app.regensbot.de/mcp
Create an accountSign in

Six clients connect through one URL to 25 tools in 6 groups.

Two ways to connect

Two ways to connect. Both at once.

ASign in

MCP sign-in with OAuth 2.1

For clients that support MCP sign-in. Add the URL, and the client opens a Regensbot page where you sign in, choose an organisation and allow the connection.

  1. Add the URL
  2. Sign in
  3. Choose an organisation
  4. Allow

Works with

  • Claude
  • ChatGPT
  • Claude Code
  • VS Code
  • MCP Inspector

BAPI key

An API key from the dashboard

For every other client, or when you prefer a key. Create a key in the dashboard and send it with every request as a header.

Request header
Authorization: Bearer rb_live_...
Open the dashboard

Works with

  • Claude Code
  • VS Code
  • Cursor
  • MCP Inspector

Both work at the same time.

Step by step

Set up your client.

01Web, Desktop and mobile

Claude

Sign in (OAuth)

  1. 01Add a custom connector and paste the MCP URL.
  2. 02Sign in to Regensbot, choose an organisation and allow the connection.
MCP URL
https://app.regensbot.de/mcp

NoteClaude identifies itself with a client metadata document, so the consent page shows it as a known app.

02Developer mode

ChatGPT

Sign in (OAuth)

  1. 01Turn on developer mode.
  2. 02Add the MCP URL as a connector.
  3. 03Sign in to Regensbot and allow the connection.
MCP URL
https://app.regensbot.de/mcp

NoteChatGPT also identifies itself with a client metadata document.

03In your terminal

Claude Code

Sign in (OAuth)API key

With sign-in

Run this command. Claude Code opens the browser for the sign-in.

Terminal
claude mcp add --transport http regensbot https://app.regensbot.de/mcp

With an API key

Or connect with an API key in the header:

Terminal
claude mcp add --transport http regensbot https://app.regensbot.de/mcp --header "Authorization: Bearer rb_live_..."

Replace rb_live_... with a key from the dashboard.

04.vscode/mcp.json

VS Code

Sign in (OAuth)API key

With sign-in

Add a server of type http with the MCP URL. VS Code registers itself and signs in.

.vscode/mcp.json
{
  "servers": {
    "regensbot": {
      "type": "http",
      "url": "https://app.regensbot.de/mcp"
    }
  }
}

With an API key

Or send an API key in the headers:

.vscode/mcp.json
{
  "servers": {
    "regensbot": {
      "type": "http",
      "url": "https://app.regensbot.de/mcp",
      "headers": {
        "Authorization": "Bearer rb_live_..."
      }
    }
  }
}

Replace rb_live_... with a key from the dashboard.

05~/.cursor/mcp.json

Cursor

API key

With an API key

Add the MCP URL and the key header under "url" and "headers".

~/.cursor/mcp.json
{
  "mcpServers": {
    "regensbot": {
      "url": "https://app.regensbot.de/mcp",
      "headers": {
        "Authorization": "Bearer rb_live_..."
      }
    }
  }
}

Replace rb_live_... with a key from the dashboard.

NoteCursor connects with an API key. Its sign-in callback uses a cursor:// address, which Regensbot's client registration does not accept.

06Any MCP client

MCP Inspector and other clients

Sign in (OAuth)API key

  1. 01Connect to the MCP URL.
  2. 02Sign in through Dynamic Client Registration, or send the key header.
MCP URL
https://app.regensbot.de/mcp
Request header
Authorization: Bearer rb_live_...

Replace rb_live_... with a key from the dashboard.

The tools

25 tools. One runs the audit.

Your assistant picks the tool. audit_url runs an audit and counts against your monthly audits. Crawls, renders and re-checks run within your plan's limits, and the knowledge-base tools have no monthly quota.

Audit05

Run an audit and read its findings.

01audit_url
Audits one URL against the 138 requirements: one finding per requirement, each with its evidence and sources, and a run ID.Counts as an audit
02get_audit
Gets one audit run again, with the findings that arrived later merged in: PageSpeed Insights, renders and Claude's proposals. Crawl findings come from get_site_findings.
03get_finding
Gets one finding of an audit.
04list_recent_audits
Your organisation's recent audits, with their status and counts.
05get_usage
Your plan, audits used and left, the reset date and a billing link.

Knowledge base05

Look up requirements, claims and their evidence.

06get_requirement
Gets one requirement of the developer kit.
07list_requirements
Lists the requirements of the developer kit.
08search_knowledge_base
Searches the knowledge base.
09get_claims
Gets claims from the knowledge base by their IDs.
10get_claim_evidence
Shows the evidence behind a claim.

Sites and checklist06

Your verified sites, their checklist and their evidence.

11list_sites
Lists your organisation's sites.
12get_site
Gets one site by its domain or ID, with its hosts and profile.
13get_checklist
The site's checklist: the state of each requirement, with filters for state and area.
14get_check_guide
The guide for one requirement: why it matters, the steps and the evidence it needs.
15get_coverage
The site's coverage of the 138: automatic, assisted or guided, and the source each one still needs.
16request_upload_link
A single-use link to upload evidence in the dashboard. Files never travel through MCP.

Workflows05

Drills, removals, site moves and re-checks after a fix.

17start_maintenance_drill
Schedules a maintenance drill: during the window, Regensbot probes the site and mails the result.
18open_removal_incident
Opens an urgent removal: checks that the URLs send a lasting removal signal and re-checks them until it holds.
19register_migration
Registers a site move with its redirect map, and re-checks the redirects on a schedule.
20mark_fixed
Marks a requirement as fixed; Regensbot re-checks it after Google's processing window.
21get_job
The status of a job that runs in the background.

Crawl03

Crawl a verified site and read its site-level findings.

22start_site_crawl
Starts a crawl of a verified site and returns its crawl ID.
23get_crawl_status
A crawl's progress, estimated time left and completeness.
24get_site_findings
The site-level findings of a crawl, each with its scope.

Server logs01

Verified crawler traffic from your server logs.

25get_crawler_traffic
Verified crawler traffic from your server logs, by status, template and crawler type.
Limits

Limits your assistant can read.

  • 01

    Counted first

    Every audit is counted against your organisation's plan before anything is fetched.

  • 02

    Not counted

    Refused addresses and sites that could not be reached do not count.

  • 03

    A readable refusal

    A refusal, such as the monthly quota, the hourly limit or a refused address, is a tool error the assistant can read. Never an HTTP error.

  • 04

    Yours only

    Each organisation sees only its own audits and sites.

Plan limits
Tool result · audit_urlDemo data

isErrortrue

content[0].text

“Monthly quota reached on the Free plan. It resets on 2026-11-01 (UTC).”

The assistant reads it

Your Free plan has no audits left this month. They reset on 1 November. get_usage has the billing link if you want more.

Create an account, then connect.

Sign up, confirm your address, and add the URL to your client.