Authorix MCP Server is a single-file 📝Model Context Protocol (MCP) server that exposes the 📝Authority Index to AI clients as four tools, letting an assistant look up ranked human experts mid-conversation.

It is the agent-facing wrapper around the 📝Authorix API, mapping each REST endpoint to one tool with the same semantics and the same credentials. It runs as a local stdio subprocess rather than a hosted endpoint, so the API key stays on the machine and is never transmitted to the model provider or any third party. The practical use is grounding: an assistant answering "who should I cite on this" or "is this account actually an authority" can query indexed entities instead of recalling them, which is the difference between a sourced answer and a plausible one.

Prerequisites

• Python 3 — The server ships as a single server.py; some systems need the absolute interpreter path from which python3.
• Packages — pip install mcp requests covers both dependencies.
• API key — The same organization key used for the REST API, supplied through the AUTHORIX_API_KEY environment variable.
• Positive balance — Tool calls bill against the organization's balance exactly as direct API calls do.

Interface

Install the server, then register it with a client:

curl -o server.py https://authorix.com/mcp/server.py
pip install mcp requests

search

Searches the Authority Index for a topic and returns the ranked top authorities in that niche.

Parameters:

• query (string, required) — the topic, keyword, or niche

Returns: a ranked list of authorities with names, ranks, and topics.

detail

Retrieves the full profile for one authority by its Authority Index identifier.

Parameters:

• sdu (integer, required) — the authority identifier, e.g. 294152

Returns: bio, authority score, social handles with follower counts, and website.

match

Resolves a social handle to a ranked profile, answering whether the account belongs to a recognized authority.

Parameters:

• handle (string, required) — the social media handle to resolve

Returns: the matched profile with name, rank, country, and core topic.

peers

Maps the influence circle around a handle.

Parameters:

• handle (string, required) — the handle whose network to map

Returns: the top authorities connected to that account.

Code Examples

📝Claude Desktop, in ~/Library/Application Support/Claude/claude_desktop_config.json:

{
  "mcpServers": {
    "authorix": {
      "command": "python3",
      "args": ["/path/to/server.py"],
      "env": { "AUTHORIX_API_KEY": "your_api_key_here" }
    }
  }
}

📝Claude Code, either as a one-liner or in a project-scoped .mcp.json:

claude mcp add authorix -- python3 /path/to/server.py

📝Cursor uses the same object shape in .cursor/mcp.json.

Error Handling

• Server fails to start — Usually a python3 not on PATH in the client's environment; replace command with the absolute path from which python3.
• Tools appear but every call fails — The AUTHORIX_API_KEY env var did not reach the subprocess. Client config env blocks do not inherit your shell.
• 403 on every tool — Either an invalid key or an exhausted balance; both surface through the underlying REST API and are resolved from the Authorix dashboard.

Related

• 📝Authorix — the platform this server exposes, including its consumer surfaces.

Contexts

• 🏷️#authorix
• 🏷️#feature-of-authorix