Search for OpenTelemetry traces with filters.

Supports both simple parameters and advanced generic filter system.

Get complete trace details by trace ID.

Returns all spans with attributes, including parsed Opentelemetry data for LLM operations.

Synthesize a likely-root-cause diagnosis for a trace, instead of returning raw trace data for the caller to re-derive one from every time.

Computes a critical path (the "Last Finishing Child" chain actually responsible for the trace's total latency), ranks spans by self-time (latency contribution net of children, top 10), and - when the trace contains an error anywhere under any root span - identifies the deepest error span in the trace's error chain as the likely root cause. Falls back to the highest self-time span as a pure-latency diagnosis when no error is present. Deterministic (no LLM call); works against any configured backend.

Try to find the corresponding trace in a second, independently- configured backend (e.g. a Datadog trace and its downstream Sentry error, joined) - given a trace_id known to the primary backend.

Tries a direct trace_id match in the secondary backend first (confidence "high"); if that fails, falls back to a time-window + service-name-overlap heuristic search (confidence "low"). This is a best-effort correlation, not a guaranteed join - see the always-present limitations in the result for why. Requires a secondary backend to be configured via SECONDARY_BACKEND_TYPE/SECONDARY_BACKEND_URL (and any backend-specific fields) environment variables; raises a clear error otherwise.

Get aggregated LLM usage metrics (token counts) for a time period.

Provides breakdowns by model and service.

List all available services in the OpenTelemetry backend.

Returns: JSON string with list of services

Find traces with errors.

Including detailed error messages, stack traces, and LLM-specific error information.

List all LLM models being used with usage statistics.

Discovers what models are deployed and tracks their usage patterns.

Get detailed performance statistics for a specific LLM model.

Analyzes request count, latency percentiles (p50, p95, p99), token usage statistics, error rates, and finish reason distributions.

List conversations/sessions grouped by gen_ai.conversation.id.

Groups spans that carry the gen_ai.conversation.id attribute (a real, cross-industry OTel semantic convention for session/conversation grouping) to surface per-conversation span counts, token usage, and time bounds - useful for understanding multi-turn conversation activity.

Get detailed statistics for a single conversation/session.

Analyzes span count, distinct services, time bounds, LLM request/success/ error counts, latency percentiles, and token usage for every span sharing the given gen_ai.conversation.id.

Compare aggregated LLM usage metrics between two time windows.

Runs the same usage aggregation for both ranges and returns the delta - useful for "this week vs last week" or "before/after a deploy" style comparisons of request/token counts.

Investigate an LLM cost spike: compare a recent window against a baseline and rank which models/services contributed most to the change.

On-request/pull-based analysis, not a push alert - mirrors SigNoz's own "investigate telemetry cost" skill. Call this when you suspect (or want to check for) a cost increase, rather than polling get_llm_usage by hand.

Investigate an error-rate spike: compare a recent window against a baseline and rank which services/models/error types contributed most.

is_spike requires both an absolute error-count floor and a relative rate-multiplier to hold, so a tiny sample (e.g. 1 error becoming 2) doesn't read as a spike.

Get aggregated performance stats grouped by prompt name and version.

Groups spans by gen_ai.prompt.name + gen_ai.prompt.version, mirroring Langfuse's shipped per-prompt Metrics tab. Real-world adoption of these two attributes is still thin, so this tool may often return an empty list until more instrumentations populate them.

Find traces with highest LLM token usage.

Useful for cost optimization and identifying inefficient prompts.

Find slowest LLM traces by duration.

Useful for performance optimization and identifying latency bottlenecks.

Search for individual OpenTelemetry spans with optional filters.

Unlike search_traces, this returns individual spans rather than grouped traces, which is useful for analyzing specific operations or finding spans with certain characteristics (e.g., LLM tool calls with traceloop.span.kind == tool).

List all LLM tools being used by identifying traceloop.span.kind == tool.

Discovers which tools/functions LLM applications are calling, grouped by tool name with usage statistics.