
Trace Forge
Gives Claude write access to create structured execution traces with spans, tool calls, and decision logging. You get six tools: trace_start to begin tracking an agent task, trace_span for nested work units, trace_tool_call to log each tool invocation with duration and cost estimation, trace_decision to record reasoning and alternatives considered, trace_end to finalize with automatic metrics, and get_trace plus search_traces for retrieval. Stores everything locally via stdio transport. Built specifically for EU AI Act Article 14 compliance, which mandates traceability for high-risk AI systems starting August 2026. The only observability MCP that writes data instead of just reading it.
- tools
- 7
- transport
- STDIO
- last updated
- Apr 24, 2026
Tools
7 toolsPublic tool metadata for what this MCP can expose to an agent.
trace_startBegin a new trace for an agent task.
Begin a new trace for an agent task.
No parameter schema in public metadata yet.
trace_spanCreate a span within a trace. Spans can be nested via `parent_span_id`.
Create a span within a trace. Spans can be nested via `parent_span_id`.
No parameter schema in public metadata yet.
trace_tool_callLog a tool call within a trace. Auto-calculates cost estimate from token count.
Log a tool call within a trace. Auto-calculates cost estimate from token count.
No parameter schema in public metadata yet.
trace_decisionLog an agent decision point with reasoning and alternatives considered.
Log an agent decision point with reasoning and alternatives considered.
No parameter schema in public metadata yet.
trace_endComplete a trace. Auto-calculates total duration, event counts, and span counts.
Complete a trace. Auto-calculates total duration, event counts, and span counts.
No parameter schema in public metadata yet.
get_traceRetrieve a complete trace tree with all spans, events, tool calls, and decisions.
Retrieve a complete trace tree with all spans, events, tool calls, and decisions.
No parameter schema in public metadata yet.
search_tracesFind traces by agent ID and/or status.
Find traces by agent ID and/or status.
No parameter schema in public metadata yet.
Sponsors
Readme
trace-forge-mcp
MCP server for writing structured traces, spans, and decisions. The only write-path observability MCP — all 9 existing observability MCPs are read-only.
Built for EU AI Act traceability compliance (Article 14, due August 2026).
Install
npx trace-forge-mcp
Claude Desktop
Add to claude_desktop_config.json:
{
"mcpServers": {
"trace-forge": {
"command": "npx",
"args": ["trace-forge-mcp"]
}
}
}
From source
git clone https://github.com/mdfifty50-boop/trace-forge-mcp.git
cd trace-forge-mcp
npm install
node src/index.js
Tools
trace_start
Begin a new trace for an agent task.
| Param | Type | Required | Description |
|---|---|---|---|
agent_id | string | yes | Unique agent identifier |
task_description | string | yes | What the agent is doing |
metadata | object | no | Optional metadata (model, department, priority) |
Returns: { trace_id, started_at }
trace_span
Create a span within a trace. Spans can be nested via parent_span_id.
Details
- Repository
- github.com/mdfifty50-boop/trace-forge-mcp
- Category
- Monitoring & Observability
- Package
trace-forge-mcp- Transport
- STDIO
- Registry
- MCP registry, active