A side-by-side look at the MCP Trace and MCP Agent Trace Inspector MCP servers: transport, auth, tools, quality signals, and the config to install either one in Claude, Cursor, Windsurf or VS Code.
Verdict: choose MCP Trace for monitoring work; choose MCP Agent Trace Inspector for monitoring as a local process. You can also run both side by side in one mcpServers block.
list_runs, run_summary, span_tree.trace_start, trace_step, trace_end.Rows where the two servers differ are marked.
| Spec | M MCP Trace | |
|---|---|---|
| Category | Monitoring | Monitoring |
| Quality signal | 47/100 · Fair | 49/100 · Fair |
| Install command | Not confirmed yet | Confirmed |
| Transport | — | Local subprocess (stdio) |
| Auth | No auth required | Not declared |
| Pricing | Free | Not declared |
| Required env vars | None listed | None listed |
| Verified owner | Community listing | Community listing |
| GitHub stars | 0 | 0 |
| Directory activity | 0 views 0 installs 0 upvotes | 1 views 0 installs 0 upvotes |
| Tools | 12 | 13 |
| Full listing | View MCP Trace | View MCP Agent Trace Inspector |
list_runsStarting out — recent runs with task, model, duration, cost, decision countsrun_summaryOne run at a glance (accepts trace-id prefix)span_tree"What did the agent actually do?" — nested shape of the runslowest_spans"Why was it slow?" — top-k spans by durationapproval_logHITL audit — every approve/deny/**edit**, who decided, and *the rationale*token_usageCost questions — aggregated across runs or per-runsearch_spansFind spans by tool name, file path, "denied", …failure_reportFind actionable, bounded, redacted errors and failed tool calls.tool_statsRank tools by volume, error rate, denials, and latency.security_auditAudit caller identity, tenant, server provenance, risk, and approval evidence.recent_activityPoll new spans with a cursor while a run is active.compare_runsCompare selected traces for regressions across models or versions.trace_startbegin a new trace; returns a `trace_id` for subsequent callstrace_steprecord one tool call step (inputs, outputs, optional token count and latency)trace_endmark a trace as completedlist_traceslist stored traces with names, statuses, and timestampsget_trace_summarytoken totals, step count, latency, and cost estimate for a tracecompare_tracesdiff two traces side by side (step counts, tokens, latency)extract_reasoning_chainextract only reasoning/thinking steps from a traceexport_dashboardgenerate a self-contained single-file HTML dashboard with latency waterfallexport_otelexport one or all traces in OpenTelemetry OTLP JSON span formatexport_compliance_logexport the compliance audit log as JSON or CSV, with optional date range filteringconfigure_alertsconfigure alert rules on latency, error rate, or cost; fire to Slack or generic webhooksset_retention_policyset how many days to keep traces (in-memory; must be called before `apply_retention`)Paste either block (or both) into your client config, e.g. claude_desktop_config.json or ~/.cursor/mcp.json. We only show a config when the install command is confirmed.
No confirmed install command for MCP Trace yet — follow the setup steps linked from its listing page rather than a guessed package name.
{
"mcpServers": {
"mcp-agent-trace-inspector": {
"command": "npx",
"args": [
"-y",
"mcp-agent-trace-inspector@latest"
]
}
}
}