/docs
obsrv documentation
Reliability and observability for AI runtimes — agents, multimodal systems, and computer-use workflows. This is everything you need to instrument, query, and operate obsrv.
Feature map
→Every product surface and the end-to-end runtime flow.
Quickstart
→Send your first trace in under three minutes.
Python SDK
→Idiomatic context managers, async ingest, fail-soft defaults.
Node SDK
→Symmetrical TypeScript API with AsyncLocalStorage parenting.
REST API
→Every endpoint, every payload, every error code.
Trace schema
→The canonical contract between SDKs, the API, and the dashboard.
Self-host & operate
→Run obsrv in your own VPC. Architecture, env, and migrations.
What obsrv is
obsrv is a full-stack observability platform for AI agents, LLM applications, and multimodal systems. Customers install one of our SDKs in their agent code; traces flow through a Go ingest API into tenant-scoped object storage and a search-indexed warehouse; the dashboard renders the runs back as inspectable timelines, evaluations, and clusters.
These docs describe the system as it ships today. Where features are in early access, we say so explicitly.
Where to start
- Read the quickstart and send your first trace.
- Skim Concepts to learn the data model.
- Pick the Python or Node SDK and instrument your agent.
- Add an OpenAI or Anthropic adapter for automatic LLM-step capture.
- Visit the dashboard and use saved filters and clusters to explore your traces.
Help & support
- Hobby — community Slack and GitHub Discussions.
- Pro — Slack support with engineering on call.
- Enterprise — dedicated channel with an SLA.