MCPJam Inspector

What is MCPJam Inspector about?

MCPJam Inspector provides a local‑first UI for interacting with MCP servers. It lets developers observe JSON‑RPC traffic, experiment with tools and resources, and run LLM chats against their server implementations.

How to use MCPJam Inspector?

# Quick start via npx (recommended)
npx @mcpjam/inspector@latest

Optional flags:

--port <number> – run on a custom port.
--ollama <model> – start with an Ollama model.
--config mcp.json – import server definitions from a mcp.json file.

The UI is then reachable at http://localhost:3001 (or the custom port you chose). You can add servers via the MCP Servers tab, selecting STDIO, SSE, or HTTP transports, and test OAuth 2.0 or Dynamic Client Registration.

Key features of MCPJam Inspector

Protocol handshake testing – visual verification of tools, resources, prompts, elicitation, and OAuth.
All transports – STDIO, Server‑Sent Events, and streamable HTTP.
LLM Playground – chat with OpenAI, Anthropic Claude, Google Gemini, Ollama, etc., to see server behavior against real models.
OAuth testing – full flow for OAuth 2.0 and dynamic client registration.
JSON‑RPC viewer – see every request/response payload for granular debugging.
MCP‑UI & OpenAI Apps SDK support – deterministic tool calls and UI rendering in the Apps SDK context.
Docker deployment – run the inspector in a container (docker run -p 3001:3001 mcpjam/mcp-inspector:latest).

Use cases of MCPJam Inspector

Debugging custom MCP servers – step through handshakes, tool calls, and prompt generation.
Testing AI integrations – verify how your server interacts with different LLM providers.
OAuth implementation validation – ensure correct token exchange and client registration.
Rapid prototyping – iterate on tool definitions and UI without redeploying the server.
Educational demos – showcase MCP capabilities to teams or stakeholders.

FAQ from the MCPJam Inspector

Q: Do I need to install Node.js? A: Yes, Node.js 20+ is required. The tool runs via npx so no global install is needed.

Q: Can I run the inspector offline? A: The UI itself works offline, but LLM playground features need internet access to contact external model APIs unless you use a local Ollama model.

Q: How do I add a server that uses STDIO? A: Use the Add Server UI, select STDIO, and provide the command (e.g., uv run fastmcp run path/to/server.py). The inspector will launch the process and communicate via pipes.

Q: Is there support for Windows/macOS desktop apps? A: Pre‑built desktop binaries are available on the MCPJam website; they internally launch the same npx‑based server.

Q: Where can I report bugs or request features? A: Open an issue on the GitHub repository or join the Discord community linked in the README.

www.mcpjam.com

MCPJam inspector is the testing and debugging platform for MCP servers & OpenAI apps. Visually inspect your server's tools, resources, prompts, and OAuth. Try your server against different models in the LLM playground.

🚀 Quick Start

Start up the MCPJam inspector:

npx @mcpjam/inspector@latest

Installation Guides
Key Features
Contributing
Links
Community
Shoutouts
License

Installation Guides

Requirements

Install via NPM

We recommend starting MCPJam inspector via npx:

npx @mcpjam/inspector@latest

We also have a Mac and Windows desktop app:

Docker

Run MCPJam Inspector using Docker:

# Using Docker Compose (recommended)
docker-compose up -d

# Or using Docker run
docker run -d \
  -p 6274:6274 \
  --env-file .env.production \
  -e NODE_ENV=production \
  --add-host host.docker.internal:host-gateway \
  --name mcp-inspector \
  --restart unless-stopped \
  mcpjam/mcp-inspector:latest

The application will be available at http://127.0.0.1:6274.

Important for macOS/Windows users:

Access the app via http://127.0.0.1:6274 (not localhost)
When connecting to MCP servers on your host machine, use http://host.docker.internal:PORT instead of http://127.0.0.1:PORT

Example:

# Your MCP server runs on host at: http://127.0.0.1:8080/mcp
# In Docker, configure it as: http://host.docker.internal:8080/mcp

Key features

Capability	Description
Multi-protocol servers	Connect to STDIO, SSE, and streamable HTTP MCP servers.
Flexible auth	Supports OAuth 2.1 and bearer tokens, including custom scopes and client credentials.
Rich configuration	Configure environment variables, custom headers, and timeouts.
Manual tool invocation	Manually invoke MCP tools, resources, resource templates, and elicitation flows.
Server info	View server icons, version, capabilities, instructions, and ChatGPT widget metadata exposed by the server.

OpenAI Apps & MCP-UI

Develop OpenAI apps or MCP-UI apps locally. No ngrok needed. MCPJam is the only local-first OpenAI app emulator.

OAuth Debugger

View every step of the OAuth handshake in detail, with guided explanations.

LLM Playground

Try your server against any LLM model. We provide frontier models like GPT-5, Claude Sonnet, Gemini 2.5. No API key needed, it's on us.

Contributing 👨‍💻

We're grateful for you considering contributing to MCPJam. Please read our contributing guide.

You can also reach out to the contributors that hang out in our Discord channel.

Links 🔗

Roadmap (TBD)
Website
Blog
Pricing
Docs

Community 🌍

Shoutouts 📣

Some of our partners and favorite frameworks:

Stytch - Our favorite MCP OAuth provider
DooiLabs/FastApps - The Python framework to build OpenAI Apps.
xMCP - The Typescript MCP framework. Ship on Vercel instantly.
Alpic - Host MCP servers

License 📄

This project is licensed under the Apache License 2.0 - see the LICENSE.