Redact spans with your own service
Call a detection service you run when TypeScript spans are exported.
Everything we have shipped lately: new features, improvements, and fixes.
20 updates
Call a detection service you run when TypeScript spans are exported.
The TypeScript SDK now traces TypeSafe AI's Jev calls as decision spans with model, tokens and cost.
Detectors now group findings into signals: one per recurring problem, each with one root cause analysis.
List evaluations, runs and dataset versions from the CLI, or ask the assistant how a run scored.
Power your detectors with Jev using your own TypeSafe AI key.
Run read-only SQL over your project's spans and traces from the CLI or the API.
Build versioned datasets from production spans, evaluation runs or your SDK.
Run your agent over a dataset from Python or TypeScript, then review scores, traces and run comparisons in TraceRoot.
traceroot dashboards bring agent latency, cost, tokens, and failures into configurable views, with a widget builder and links to the underlying traces.
List recent traces, inspect a span tree and export trace context from the CLI.
Follow each conversation turn through its model calls, tools and nested agents.
Automatic root cause analysis now uses the BYOK model and provider saved for your project.
Trace OpenRouter requests through the OpenAI-compatible SDK in Python or TypeScript.
Inspect input, output and cache usage alongside the cost of a trace or individual span.
Capture Microsoft Agent Framework runs, model calls and tools in traceroot.
Capture Pydantic AI agent runs, model calls and tools with the Python SDK.
Inspect model calls and tool steps from Vercel AI SDK applications in traceroot.
Trace agent runs, model calls and tools across four more Python integrations.
Find slow steps, parallel calls and gaps with traceroot's new Timeline view.
Open a running agent trace in traceroot to see completed model calls and tool steps as they arrive, without refreshing the page.