inspector

The MCPJam Inspector is a local development tool that lets you test and debug your MCP servers, tools, resources, prompts, and OAuth flows with a full JSON-RPC observable console. It provides an emulator for ChatGPT apps and MCP apps, a UI for inspecting server capabilities, and a playground to invoke tools or flows against your own MCP server. Use it to connect to your running MCP servers locally, run tool invocations, view logs, and verify how your server handles different RPC messages and responses. The inspector supports all protocol versions and offers an integrated workflow for debugging OAuth, resources, and elicitation tools, making it easier to iterate on MCP-based apps and servers.

Prerequisites:

• Node.js (recommended latest LTS, 12+ or higher depending on your environment)
• npm or pnpm

Option A: Run directly with npx (no install required)

• Ensure you have Node.js installed
• Run:

npx @mcpjam/inspector@latest

Option B: Install locally via npm (if you prefer a pinned version)

• Install the package globally or add to your project:

npm install -g @mcpjam/inspector@latest

Then run the binary or script as documented by the package (follow the CLI help if needed).

Option C: Docker (alternative deployment)

• Run the inspector in a container (if provided by MCPJam) with proper port mapping:

docker run -p 6274:6274 mcpjam/mcp-inspector

Prerequisites for Docker: Docker installed and running, appropriate permissions to pull images.

Tips and common issues:

• If you’re using npx frequently, the -y flag in the mcp_config example is optional; you can omit it if you prefer sticking with the default npx behavior.
• The inspector exposes a local UI. Make sure you expose the correct port (default 6274) if you’re running behind a NAT or in a container.
• When debugging OAuth or tools, check the API version compatibility and ensure your MCP server responds with the expected JSON-RPC payloads.
• If you upgrade your MCPJam Inspector, verify that your server’s RPC schemas remain compatible with the inspector’s observer tooling.
• Environment variables can be used to configure endpoints, tokens, or feature flags; add them under the env field in the mcp_config if your deployment requires them.