AI assistants and MCP
Let Claude, ChatGPT, Cursor and other AI tools search the catalogue and run calculations with your key.
The 96Astro API runs a hosted Model Context Protocol (MCP) server. Connect it to an AI assistant or coding agent and it can find the right calculation, read its exact input and call it for you, with your own key.
https://api.96astroapi.com/mcpIt uses the Streamable HTTP transport. Your key goes in the Authorization header, as with the REST API.
Start with a sandbox key (sk_test_): sandbox calls are free. With a live key, every calculation the assistant runs spends credits, exactly as a direct API call would.
Tools
| Tool | What it does | Key |
|---|---|---|
search_capabilities | Finds calculations by keyword, e.g. "manglik" or "choghadiya" | Not needed |
list_families | Lists the families of calculations | Not needed |
describe_capability | The exact input schema, an example, options, cost and languages of one capability | Not needed |
call_capability | Runs one capability and returns its JSON | Your key |
find_place | Turns a place name into latitude, longitude and time zone | Your key |
call_capability goes through the same /v1 endpoint as your own code, so credits, rate limits, idempotency and errors work the same way. Chart images are left out of the tool result to save the model's context; call the API directly when you need them.
Claude Code
claude mcp add --transport http 96astro https://api.96astroapi.com/mcp \
--header "Authorization: Bearer sk_test_..."Then ask, for example: "Using 96astro, what is the moon sign for someone born 18 October 1990 at 14:30 in Delhi?"
Cursor
In .cursor/mcp.json (this project) or ~/.cursor/mcp.json (every project):
{
"mcpServers": {
"96astro": {
"url": "https://api.96astroapi.com/mcp",
"headers": { "Authorization": "Bearer sk_test_..." }
}
}
}VS Code
In .vscode/mcp.json. VS Code asks for the key once and stores it securely:
{
"inputs": [{ "type": "promptString", "id": "astro-key", "description": "96Astro API key", "password": true }],
"servers": {
"96astro": {
"type": "http",
"url": "https://api.96astroapi.com/mcp",
"headers": { "Authorization": "Bearer ${input:astro-key}" }
}
}
}Claude Desktop and other clients without remote servers
A client that only starts local (stdio) servers can reach it through the mcp-remote bridge, which needs Node.js:
{
"mcpServers": {
"96astro": {
"command": "npx",
"args": ["-y", "mcp-remote", "https://api.96astroapi.com/mcp", "--header", "Authorization:${ASTRO_AUTH}"],
"env": { "ASTRO_AUTH": "Bearer sk_test_..." }
}
}
}Any other client that supports Streamable HTTP with a custom header works the same way: the URL above and Authorization: Bearer <key>.
Docs for language models
Without MCP, point an assistant at the documentation written for it:
- /llms.txt: what the API is, how to call it, and a link to every guide and capability.
- /llms-full.txt: every guide and the whole catalogue in one file.
- Any guide or capability page as Markdown: add
.mdto its address, e.g. /docs/quickstart.md. Each guide also has Copy Markdown and Open buttons under its title.
Keep your key safe
- Keep MCP configuration files with a key out of version control, or use the client's secret prompt (as in the VS Code example).
- The key can do whatever your account can; give an assistant a sandbox key unless it needs live data.
- Revoke a key under Keys in the dashboard at any time; the connection stops working at once.