mcpsnoop-mcp-debugging

Inspect and replay MCP client-server JSON-RPC traffic through a transparent proxy.

7|2|Updated May 16, 2026
One-click install
npx skills add https://github.com/reason-machines/mcp-skills --skill mcpsnoop-mcp-debugging-reason-machines
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: mcpsnoop-mcp-debugging
Source: https://github.com/reason-machines/mcp-skills/tree/main/skills/mcpsnoop-mcp-debugging
Command: npx skills add https://github.com/reason-machines/mcp-skills --skill mcpsnoop-mcp-debugging-reason-machines

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Debugging MCP servers is difficult because the communication between AI clients and servers is invisible, making it hard to diagnose why tools are not called, calls hang, or capabilities mismatch. ## Core Features & Use Cases - Live Traffic Interception: Wrap any MCP server command with mcpsnoop to capture every JSON-RPC frame between your real client and server in an interactive TUI. - Filtering and Inspection: Filter frames by tool, method, direction, kind, and status, then drill into full request and response payloads. - Tool Call Replay: Replay any captured tool call against a fresh server instance to test fixes without restarting your client. - Use Case: Your Claude Desktop filesystem tool silently fails. Wrap the server with mcpsnoop, open the TUI, press 'c' to verify capability negotiation, filter with /tool:read_file, and discover the call hangs in PENDING state due to a permission error visible in stderr. ## Quick Start Wrap my MCP server launch command with mcpsnoop in my client config and open the mcpsnoop TUI to inspect the live traffic.

Frequently Asked Questions about mcpsnoop-mcp-debugging

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I debug MCP server traffic between my client and server?▼

Wrap your MCP server command with mcpsnoop in your client configuration, placing your normal launch command after the -- separator. Then run mcpsnoop to open the TUI, which displays every JSON-RPC frame exchanged between the real client and server.

What is the difference between mcpsnoop and MCP Inspector?▼

MCP Inspector connects to your server as a separate client, while mcpsnoop acts as a transparent proxy intercepting the actual traffic between your real AI client and server. This shows genuine client behavior rather than simulated requests.

Does mcpsnoop support HTTP MCP servers?▼

Yes, mcpsnoop supports streamable HTTP servers through reverse proxy mode. Run mcpsnoop http with --target pointing to your server and --listen for the proxy port, then point your client to the proxy address.

Why is my MCP tool call stuck in PENDING status?▼

A PENDING status means the server received the request but never responded, usually due to resource locks, external API timeouts, or deadlocks. Check the server stderr frames in the TUI and press enter on the pending frame to inspect the full request.

How do I replay an MCP tool call for testing?▼

In the mcpsnoop TUI, navigate to a tools/call request frame and press r. mcpsnoop spawns a fresh isolated server instance, executes the identical call, and displays the new response, letting you test fixes without restarting your client.