Connect your AI assistant to Selfcare

Claude

1. In Settings → Connectors, choose Add custom connector.
2. Give it a name (Selfcare) and paste the URL https://api.selfcare.work/mcp.
3. Click Connect. The Selfcare authorisation screen opens: sign in with your account, choose the company and confirm what the assistant can do.

Claude Code

1. In the terminal, add the server with the command below.
2. Inside Claude Code, run /mcp and choose the selfcare server. The browser opens so you can sign in to Selfcare and authorise.

claude mcp add --transport http selfcare https://api.selfcare.work/mcp

ChatGPT

1. In Settings → Apps and connectors, turn on developer mode (in the advanced settings) and choose Create.
2. Paste the URL https://api.selfcare.work/mcp and choose OAuth as the authentication.
3. Confirm on the Selfcare authorisation screen, as with Claude.

Cursor

1. Open the Cursor settings, MCP section, and add the server to mcp.json (project or global).
2. On first use, Cursor opens the browser so you can sign in to Selfcare and authorise.

{
  "mcpServers": {
    "selfcare": { "url": "https://api.selfcare.work/mcp" }
  }
}

Other MCP clients

Any client that speaks MCP over Streamable HTTP and does OAuth 2.1 with discovery connects the same way:

1. Configure a remote MCP server with the URL https://api.selfcare.work/mcp.
2. The client receives the 401, reads the metadata, registers itself and opens the browser on the Selfcare authorisation screen.
3. You sign in with your account, choose the company and approve the scopes. The screen warns that this is an AI assistant.

Without OAuth: API key

A client that doesn't do OAuth can use a personal API key, created in Settings → API with the scopes of the tools you'll use. The key is yours: keep it out of repositories and shared files.

1. Send the key in the Authorization header, as in the example.

{
  "mcpServers": {
    "selfcare": {
      "url": "https://api.selfcare.work/mcp",
      "headers": { "Authorization": "Bearer <your API key>" }
    }
  }
}