sigiro docs

sigiro is an observability backend built for agents instead of dashboards. It ingests OpenTelemetry and answers the two questions an automated loop has to answer before it can fix anything:

1. What changed? — sigiro anomalies returns shifts that deviate from the service's own statistical baseline, ranked, and grouped into incidents across signals.
2. Why? — sigiro investigate returns one structured evidence block per service: ranked findings, per-operation error breakdowns, per-model LLM stats, sampled spans and log patterns, and a paste-ready SQL query on every row.

Both commands are also endpoints — GET /v1/anomalies and POST /v1/investigate — which is how an agent asks the same two questions. See the API reference.

Everything is one static binary. No collector to run, no query language to learn beyond SQL.

Start here

• Quickstart — run the binary, send it telemetry, ask it a question. Five minutes.
• Instrument your code with an AI agent — hand a coding agent one prompt; it wires up OpenTelemetry for you.
• Hosted beta — skip hosting anything; ship OTLP to api.sigiro.com with a tenant key.

API reference

Every endpoint, its parameters and its response shape are generated from the same document sigiro serves at /openapi.json — so the reference can never drift from the API.

• API reference — five endpoints, each with a request playground.

An agent that can read /openapi.json needs nothing else from us to get started.