MCPJam Inspector
Know your MCP server works before your users do.
MCPJam Inspector is a local playground for your MCP server. Connect it, run its tools, compare multiple models side by side on one prompt, and read the full user → agent → server trace for every call. Open source and free, forever.
ClaudeThinking…
ChatGPTThinking…
CopilotThinking…
Open source, trusted by MCP builders
A playground for your MCP server
Multi-model playground
Run one prompt against multiple models side by side and compare how each one uses your tools.
Waterfall trace
See the user → agent → server timeline for every run, down to the timing of each individual tool call.
Tool & resource testing
Call tools, prompts, and resources against your live server and inspect the exact arguments and results.
JSON-RPC logger
Read every request and response on the wire, with direction, payload, and latency, to reproduce failures fast.
Run it in seconds
Point the inspector at any local or remote MCP server and start testing. A built-in MCPJam tunnel exposes your local server to ChatGPT, Claude, and Gemini over a secure URL, so you can test it in a real host without deploying — no ngrok needed.
# Test any local or remote MCP server
npx @mcpjam/inspector@latestFrequently asked questions
Yes. The client, Inspector, core CLI, and SDK are open source and free, forever, on GitHub at github.com/MCPJam/inspector.
Yes. A built-in MCPJam tunnel exposes your local server to those hosts over a secure URL, so you can test it in a real client without deploying it first — no ngrok needed.
Tools, prompts, and resources, with full JSON-RPC request and response logging and a waterfall trace of every run.
Ship knowing it works.
Run the inspector against your MCP server, reproduce any failure, and fix it. Free, in about two minutes.
Open source · no credit card · ~2 min