Divine API - Documentation MCP Server
A public, read-only MCP server that answers "how do I use the DivineAPI REST API". It serves the already-public developer documentation: every endpoint, its parameters, its response fields, the global auth and error rules, and real captured example responses.
This server does NOT call the astrology APIs and needs NO credentials. It is a reference, not a data source. To actually run astrology requests, use the divineapi SDK or the data MCP servers:
- Indian / Vedic:
https://mcp.divineapi.com/indian/mcp - Western:
https://mcp.divineapi.com/western/mcp - Horoscope, Tarot and Numerology:
https://mcp.divineapi.com/horoscope/mcp
Public endpoint
https://mcp.divineapi.com/docs/mcp
No API key, no auth token, no OAuth. Just connect.
Tools (5)
| Tool | What it does |
|---|---|
search_docs(query, limit=8) |
Case-insensitive keyword search across every endpoint card; returns the top matching cards (most query tokens first). |
get_endpoint(path) |
The full reference card for one exact path, or the closest match / a "did you mean..." list. |
list_endpoints(category="") |
Every path [host] line, optionally filtered to a category (Indian, Western, Numerology, PDF, Horoscope, Lifestyle, Calculators). |
get_playbook() |
The global rules: auth, error semantics per host, birth params, horoscope selectors, field formats, house systems, and what not to do. |
get_example(path) |
A real captured example response body for the path (version-agnostic slug fallback), or a pointer to the card's returns: line. |
Data sources
docs-pack.txtis fetched live at startup fromhttps://developers.divineapi.com/docs-pack.txt(10s timeout); on failure the server falls back to the bundled snapshot shipped in this repo.examples.jsonis a map ofnormalized_path -> example response body, generated from the published Postman collection bybuild_examples.py. Regenerate it with:
python3 build_examples.py "/path/to/DivineAPI_Collection_with_examples_Final Version.json"
Run locally
python -m venv .venv && source .venv/bin/activate
pip install ".[http]"
# stdio (for Claude Desktop, Cursor, etc.)
divineapi-docs-mcp
# HTTP (streamable-http on :8000)
MCP_TRANSPORT=http uvicorn server:app --host 0.0.0.0 --port 8000
Deploy
Mirrors the DivineAPI data MCPs. The container runs uvicorn server:app on port 8000; docker-compose maps host port 8004 -> 8000.
docker compose up -d --build
Documentation: https://developers.divineapi.com