Trace MCP Server

Structured execution trace and span logging for AI agents

Local serverstdio

What is the Trace MCP server?

Trace MCP server exists for a simple reason — assistants are far more useful when they can act on Trace directly instead of describing what you should do. Structured execution trace and span logging for AI agents.

What you get

MCP server for writing structured traces, spans, and decisions. The only write-path observability MCP — all 9 existing observability MCPs are read-only.

What the assistant can call

Once Trace is connected, these are the calls the assistant has available:

  • trace_tool_call — after each tool invocation
  • trace_decision — at each decision point
  • trace_start — The trace_start tool exposed by this server
  • trace_span — Create a span within a trace. Spans can be nested via parent_span_id
  • trace_end — Complete a trace. Auto-calculates total duration, event counts, and span counts
  • get_trace — Retrieve a complete trace tree with all spans, events, tool calls, and decisions
  • search_traces — Returns: Matching traces sorted by recency with summary stats

Setting it up

The server ships on npm as trace-forge-mcp, so your MCP client can launch it on demand — there is no separate build step. Add the server block to your client's configuration, restart it, and the tools register themselves.

Choosing this one

This sits in the developer tooling group, where several servers overlap in what they claim to do but differ sharply once you actually set them up. Trace's toolset — trace_tool_call, trace_decision, trace_start and 4 more — is a fair guide to whether it matches your workflow. It is maintained by mdfifty50-boop; worth a glance at recent repository activity before you build anything load-bearing on it.

This entry was verified against Trace's own documentation before publication; SyncDev keeps the directory reviewed rather than auto-generated.

Before you rely on it

  • It runs with your machine's permissions. That is convenient and also the reason to think about what you point it at before you approve a tool call.
  • MCP clients confirm each tool call by default. Leave that on until you have watched what the trace mcp server does with a few real requests.

Available tools

ToolWhat it does
trace_tool_callafter each tool invocation
trace_decisionat each decision point
trace_startThe trace_start tool exposed by this server.
trace_spanCreate a span within a trace. Spans can be nested via parent_span_id.
trace_endComplete a trace. Auto-calculates total duration, event counts, and span counts.
get_traceRetrieve a complete trace tree with all spans, events, tool calls, and decisions.
search_tracesReturns: Matching traces sorted by recency with summary stats.

How to install the Trace MCP server

{
  "mcpServers": {
    "trace-forge": {
      "command": "npx",
      "args": ["-y", "trace-forge-mcp"]
    }
  }
}

Add to claude_desktop_config.json, then restart Claude Desktop.

Example prompts to try

  • Use Trace to trace tool call.
  • Use Trace to trace decision.
  • Use Trace to trace start.

Frequently asked questions

It connects Trace to MCP-compatible AI assistants such as Claude and Cursor, exposing 7 tools (trace_tool_call, trace_decision, trace_start, and more) that the assistant can call on your behalf. Instead of copying data back and forth by hand, the assistant works with Trace directly.