MCP Doctor

diagnose why your MCP servers won't connect — in one command

npx -y github:Isco0819/mcp-doctor

Thanks for going Pro. After payment is confirmed, your license key is issued manually to the email used at checkout. Then set MCP_DOCTOR_LICENSE and run the GitHub command below.

Ever run into this?

Adding an MCP server to Claude, Cursor, or Windsurf is a coin flip. When it fails you get a silent 'failed to connect' with zero detail — wrong command, missing env var, bad path, or a server that crashes on boot. You burn 30 minutes guessing.

The fix

MCP Doctor reads your config, spawns each local stdio server, runs the initialize handshake, and reports the failure category with a next step. Pro checks supported remote endpoints too.

What it does — free

Auto-detects Claude Desktop, Cursor, Windsurf and .mcp.json configs
Spawns each local stdio server and runs the real JSON-RPC initialize handshake
Pinpoints the cause: missing env var, bad command or path, timeout, or protocol mismatch

For production & teams

MCP Doctor Pro

Free tests your local stdio servers. Pro connects to supported HTTP / Streamable-HTTP endpoints and can check them in CI.

¥1,400 / month (≈ $9)

Test remote HTTP / Streamable-HTTP servers — handshake, auth check, tool count when available & latency
CI mode — checks supported local and remote transports, then emits JUnit XML
Pinpoints remote failures: 401/403 auth, wrong endpoint, timeout, protocol mismatch
Not included today: legacy GET-based SSE and watch mode are planned for a future release

Get Pro

# key issued manually after payment confirmation
export MCP_DOCTOR_LICENSE=<key>
npx -y github:Isco0819/mcp-doctor --remote
# use --ci for JUnit output

npx -y github:Isco0819/mcp-doctor