Skip to content

For developers

Use MirageScouting in your AI assistant

One MCP server, nine read-only tools. Search companies, open roles and candidates who chose to be found, from Claude, ChatGPT or Gemini. Every result links back to the MirageScouting page it came from.

Server URL

MirageScouting MCP server

https://mcp.miragescouting.de/mcp

The install steps below show this URL wherever you need it. Sign-in uses OAuth in your browser; no API key is needed.

Install

Add MirageScouting to your assistant

Pick the assistant you use. The steps follow each vendor’s own documentation, checked on 8 October 2026.

Claude (web and desktop)Custom connector
  1. 1.Open Customize > Connectors.
  2. 2.Click “+ Add”, then “Add custom connector”.
  3. 3.Enter the name MirageScouting and the server URL https://mcp.miragescouting.de/mcp, then click “Continue”.
  4. 4.Under “Authentication”, choose “Sign in now”, then click “Add”.
  5. 5.Find MirageScouting in your connectors, click “Connect” and sign in when the browser asks.
  6. 6.In a chat, use the “+” button, open “Connectors” and switch MirageScouting on.

On Team and Enterprise plans an Owner adds the connector once under Organization settings > Connectors (“Add” > “Custom” > “Web”). Members then click “Connect” under Customize > Connectors.

Claude CodeCommand line
  1. 1.Add the server with the command below.
  2. 2.Inside Claude Code, run /mcp and follow the browser sign-in.
claude mcp add --transport http miragescouting https://mcp.miragescouting.de/mcp
ChatGPTChatGPT on the web, custom MCP server
  1. 1.Open ChatGPT Plugins, select the plus button and choose “Add custom MCP server”.
  2. 2.Enter the name MirageScouting and, under Connection, the server URL https://mcp.miragescouting.de/mcp.
  3. 3.Choose OAuth as the authentication.
  4. 4.Read the risk warning, select “I understand and want to continue”, then “Create as a plugin”.
  5. 5.In the prompt box type @ and pick MirageScouting. Naming the tool helps, for example: Use MirageScouting’s search_jobs tool to find product manager roles in Berlin.

Depending on your plan and workspace you may first need to switch on developer mode or ask a workspace admin to allow custom MCP connectors.

Gemini CLICommand line
  1. 1.Add MirageScouting to the mcpServers object in ~/.gemini/settings.json, using httpUrl.
  2. 2.Start gemini and run /mcp auth miragescouting to sign in.
{
  "mcpServers": {
    "miragescouting": {
      "httpUrl": "https://mcp.miragescouting.de/mcp"
    }
  }
}
Gemini EnterpriseCustom MCP server data store, set up by an admin
  1. 1.In the Google Cloud console open Gemini Enterprise, then Data stores, then “Create data store”. Your administrator account needs the Discovery Engine Editor role.
  2. 2.Search for “Custom MCP Server” and click “Add MCP server”.
  3. 3.Enter the server URL https://mcp.miragescouting.de/mcp and choose “OAuth 2.0”. It asks for the authorization URL, token URL, client ID and client secret. The URLs are in the server’s /.well-known/oauth-authorization-server metadata; ask us for client credentials.
  4. 4.Click “Verify Auth”, then “Continue”. Pick a location, name the connector and click “Create”.
  5. 5.Open the new data store, choose Actions > Reload custom actions, select the MirageScouting actions and click “Enable actions”. Nothing is enabled by default.

Vendors rename these screens often. If a label differs from what you see, follow the vendor’s own guide for adding a custom MCP server and use this server URL: https://mcp.miragescouting.de/mcp

Tools

Nine tools, all read-only

Your assistant calls these for you. None of them writes anything.

search_companies
Find companies that published themselves to assistants.
get_company_profile
Read one company’s published profile.
search_jobs
Find live roles by role, seniority, location and remote mode.
get_job
Read one open role in full.
search_candidates
Search candidates who opted in to AI assistants.
get_candidate_profile
Read one opted-in candidate’s restricted profile.
search_skills
Look up the exact skill names MirageScouting uses. Returns no people data.
fetch
Read the full record for an id that search returned, for ChatGPT’s deep research.

Every result carries a link back to its page on MirageScouting, so you can check it and act on it.

Skills

Search by skills

Shortlist opted-in candidates against the skills a role needs, in three steps. The first two work today. The third, a whole list of skills in one question, is not switched on yet.

  1. Step 1

    Look up the exact skill names

    Ask your assistant to check a skill name first. search_skills matches the names MirageScouting uses, so “react” becomes “React” and “JS” becomes “JavaScript”. It returns skill names only, no people.

    Example prompt

    Which skill names does MirageScouting use for react and JS?
  2. Step 2

    Search candidates by one skill and a minimum level

    Name one skill and the lowest level you accept, from 0 to 10. You get opted-in candidates who reach that level, each with a link to their MirageScouting page.

    Example prompt

    Find candidates with React at level 6 or higher.
  3. Step 3Coming soon

    Search by a whole list of skills

    Give up to 20 skills, each with its own minimum level, and choose how strict to be.

    • Best fit: candidates who meet at least a number of your skills you choose, sorted by how many they meet, then by their levels on them.
    • All: only candidates who meet every skill on the list.
    • Every result shows how many skills matched, for example “matched 12 of 16”, and which ones, at what level.
    • Each level is labelled “confirmed” when a person approved it, or “assessed” when it is an assessment result.

    Not switched on yet — we enable it after our fairness and EU AI Act review.

    Example prompt

    Find candidates who meet at least 10 of these skills at level 6 or higher: React, TypeScript, Node.js, PostgreSQL, GraphQL, Docker, Kubernetes, AWS, CI/CD, Jest, Playwright, REST APIs, System Design, Redis, Kafka, Terraform.

Results are sorted only by how many of your skills a candidate meets and their levels on them. There is no hidden score, and a person decides.

Privacy

Nothing is shared unless its owner switches it on

These rules are built into the server, not just promised on this page.

Opt-in only
A candidate is searchable only after switching it on, and it is off by default. A company publishes itself to assistants, and can hide single roles.
Never contact details
Candidates’ email, phone, address and photo are never shared. Hiring-team contact details are never returned either. To reach someone, you go through MirageScouting.
Switch it off, it is gone
A candidate can turn discoverability off at any time in their settings. Their MirageScouting page then stops resolving.
Every request is logged
Each call to the server is recorded, so access can be audited.
Create a free profile and decide for yourself