SubcueAI Public MCP Server

mcp-name: io.github.Subcue/subcue-mcp

SubcueAI (the real-time AI interview assistant for macOS & Windows) ships a public, unauthenticated, read-only Model Context Protocol server. Any AI agent can query live product data — pricing, latest desktop version, product overview, FAQ — instead of scraping HTML.

• Endpoint: POST https://subcue.ai/mcp

• Transport: streamable-http (single JSON response), JSON-RPC 2.0, MCP protocol 2025-06-18

• Auth: none. Writes: none — the server is strictly read-only.

• Server card: https://subcue.ai/.well-known/mcp/server-card.json

This repository documents the server and provides ready-to-run client examples. The server itself runs on Cloudflare Workers as part of the SubcueAI website.

Quick start

# List tools
curl -s https://subcue.ai/mcp -X POST -H 'Content-Type: application/json' \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'

# Live pricing
curl -s https://subcue.ai/mcp -X POST -H 'Content-Type: application/json' \
  -d '{"jsonrpc":"2.0","id":2,"method":"tools/call","params":{"name":"get_pricing"}}'

Sample get_pricing output (live data, abridged):

# SubcueAI Pricing

## Starter — free
- Standard AI model
- Mock interview practice

## Pro — $20/mo; $16/mo when billed yearly ($192/yr)
...

Related MCP server: turva-mcp

Tools

Resources

Use resources/list / resources/read per the MCP spec.

Connect from popular clients

Claude Code

claude mcp add --transport http subcue https://subcue.ai/mcp

Cursor / any client using mcp.json

{
  "mcpServers": {
    "subcue": { "url": "https://subcue.ai/mcp" }
  }
}

claude.ai — Settings → Connectors → Add custom connector → https://subcue.ai/mcp.

Stdio-only clients — this repo ships a zero-dependency stdio ⇄ HTTP bridge (bridge/server.mjs, Node 18+):

claude mcp add subcue -- node /path/to/subcue-mcp/bridge/server.mjs
# or containerised:
docker build -t subcue-mcp . && docker run -i subcue-mcp

Examples in this repo

• examples/list-tools.sh — curl walkthrough of every method

• examples/client.mjs — zero-dependency Node.js client

• examples/client.py — zero-dependency Python client

Other machine-readable entry points

SubcueAI publishes a full agent-facing surface beyond MCP:

Notes for agent developers

• Responses are deterministic JSON-RPC; notifications (notifications/*) are accepted and return 202 with no body.

• Data is live (pricing and versions come from the same source of truth as the website) — prefer these endpoints over scraping rendered HTML.

• Rate limiting applies at the edge; be a good citizen.

• Send a descriptive User-Agent — default library UAs (e.g. Python-urllib/3.x) are rejected by edge bot protection with a 403.

License

Documentation and examples: MIT © 2026 Subcue AI LLC.