Query your APM data.
Just talk to Claude.
Search traces, check service health, analyze LLM costs, and investigate anomalies -- directly from Claude, ChatGPT, or any MCP-compatible AI assistant.
Ask it. TraceKit answers.
11 tools your AI assistant can use to query your observability data in real time.
Search traces
"Show me the slowest requests from checkout-service" -- filter by service, duration, errors, or operation name.
Service health
"How are my services performing?" -- get latency percentiles, error rates, and throughput for every service.
LLM cost analysis
"How much am I spending on LLM calls?" -- total cost, token breakdown by model, and cost trends over time.
Alerts & anomalies
"Are there any anomalies in my backend?" -- surface firing alerts, recent anomalies, and severity levels.
Root cause analysis
"What caused this anomaly?" -- get AI-powered root cause analysis with evidence chains and confidence scores.
Live debugging snapshots
"Show me captured snapshots for auth-service" -- browse variable values and stack traces from production dynamic logging.
See it in action.
Real conversations with Claude using TraceKit's MCP connector.
Service health overview
Ask about your services and get instant health metrics, error rates, and latency data.
Trace investigation
Find slow requests, drill into span waterfalls, and identify bottlenecks in seconds.
LLM cost analysis
Break down LLM spending by model and service. Track token usage and cost trends.
Error & anomaly detection
Surface anomalies, check root cause analysis, and understand what went wrong.
Up and running in 2 minutes.
One URL, one API key, done.
Copy the MCP server URL
Your server URL is https://app.tracekit.dev/mcp -- the same for all accounts.
Add it to your AI assistant
In Claude: Settings > Connectors > Add connector. Paste the URL and click "Connect".
Enter your API key
You'll see a TraceKit authorization page. Paste your API key from the TraceKit dashboard.
Start asking questions
"How are my services performing?" -- your AI assistant now has full read access to your observability data.
MCP Server URL
Available tools
Every tool follows the open Model Context Protocol spec. Works with Claude, ChatGPT, Cursor, and any MCP client.
| Tool | Description |
|---|---|
| search_traces | Search and filter distributed traces |
| get_trace_detail | Get full span waterfall for a trace |
| list_services | List services with health metrics |
| get_service_metrics | Latency percentiles, error rate, throughput |
| get_active_alerts | Currently firing alert rules |
| search_anomalies | Recent anomalies with severity |
| list_breakpoints | Active code monitoring breakpoints |
| search_snapshots | Live debugging snapshot captures |
| search_llm_calls | Individual LLM API call details |
| get_llm_stats | Aggregated LLM cost and usage stats |
| get_rca | Root cause analysis for anomalies |
Secure by design
Authentication follows the OAuth 2.1 spec with PKCE. Your API key is never exposed to the AI model -- it stays between your browser and TraceKit. All tools are read-only. No data is modified, deleted, or sent to third parties.
Works with every MCP client
Built on the open Model Context Protocol standard by Anthropic.
Stop switching tabs to check your services
Connect TraceKit to your AI assistant and ask questions in plain English. Free tier includes 200K traces/month.