MCP Agent Trace Inspector

npm mcp-agent-trace-inspector package

Local-first, MCP-native observability for agent workflows. Every tool call, prompt transformation, latency, and token count is recorded in a local SQLite database — no cloud account, no API key, no traces leaving your machine. Built specifically for MCP rather than bolted onto a generic LLM proxy.

Tool reference | Configuration | Contributing | Troubleshooting | Design principles

Key features

• Tool call tracing: Captures inputs, outputs, latency, and token usage for every step in a workflow.
• Persistent storage: Traces survive session restarts; stored locally in SQLite with no external dependencies.
• HTML dashboard: Generates a self-contained single-file dashboard with an interactive step timeline.
• Token cost estimation: Calculates USD cost per trace using a configurable model pricing table — no API calls required.
• Trace comparison: Diff two traces side by side to measure the impact of prompt or tool changes.
• Low overhead: Adds less than 5ms per step; never becomes the bottleneck.

Why this over LangSmith / AgentOps?

If your traces contain sensitive tool outputs, proprietary prompts, or data that must stay on-device, this is the right tool. If you need cross-team trace sharing or a managed SaaS, use LangSmith.

Disclaimers

mcp-agent-trace-inspector stores tool call inputs and outputs locally in a SQLite database. Traces may contain sensitive information passed to or returned from your tools. Review trace contents before sharing dashboard exports. Traces are not automatically transmitted; optional alert webhooks are available.

Requirements

• Node.js v22.5.0 or newer.
• npm.

Getting started

Add the following config to your MCP client:

{
  "mcpServers": {
    "trace-inspector": {
      "command": "npx",
      "args": ["-y", "mcp-agent-trace-inspector@latest"]
    }
  }
}

To set a custom storage path:

{
  "mcpServers": {
    "trace-inspector": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-agent-trace-inspector@latest",
        "--db=~/traces/my-project.db"
      ]
    }
  }
}

MCP Client configuration

Amp · Claude Code · Cline · Cursor · VS Code · Windsurf · Zed

Your first prompt

Enter the following in your MCP client to verify everything is working:

Start a trace called "test-run", then list the files in the current directory, then end the trace and show me the summary.

Your client should return a summary showing step count, total tokens, and latency.

Tools

Trace lifecycle (3 tools)

• trace_start — begin a new trace; returns a trace_id for subsequent calls
• trace_step — record one tool call step (inputs, outputs, optional token count and latency)
• trace_end — mark a trace as completed

Inspection (4 tools)

• list_traces — list stored traces with names, statuses, and timestamps
• `get_trac

... View full README on GitHub