API for AI agents via Model Context Protocol

A generic LLM "knows" astrology only from text in its training data — which means hallucinated dates, coordinates, aspects. Ask "when does Saturn trine my Jupiter next?" and you get a vague phrase instead of a UTC point.

@astroway/mcp — a native Model Context Protocol server. The AI agent invokes `tool_use POST /v1/transits` and gets a real UTC timestamp to the second. 4 credits per call, <100ms latency, sub-arcsecond accuracy.

• npm install -g @astroway/mcp
• Works with Claude Desktop, Cursor, Windsurf
• Tool definitions auto-generated from OpenAPI
• Streaming JSON responses

// MCP server (npx -y @astroway/mcp) gives Claude every endpoint as a tool
import Anthropic from '@anthropic-ai/sdk';
import { spawn } from 'node:child_process';

// Spawn @astroway/mcp via npx; Claude SDK auto-discovers MCP tools
const reply = await new Anthropic().messages.create({
  model: 'claude-sonnet-4-6',
  // tools auto-loaded from running MCP server: 700+ endpoints exposed
  messages: [{ role: 'user', content: 'When does Saturn next trine my natal Jupiter?' }],
});