From one SDK call to a full release gate
Orq wraps your existing agent stack. No new frameworks, no rewrites. Four steps from zero observability to blocked regressions.
Install the SDK
Add Orq to your Python or TypeScript agent with one package install. The SDK works alongside LangChain, LlamaIndex, OpenAI Agents SDK, Anthropic SDK, and any agent you write yourself. No framework lock-in.
Instrument your agent spans
Wrap your agent entry point with one decorator or context manager. Orq automatically captures every LLM call, tool invocation, and handoff between agents, each as a named span with latency, token counts, and the raw input/output pair.
Define your eval suite
In the Orq dashboard, write metric thresholds for what good looks like: coherence score above 85%, grounding above 80%, latency under 3 seconds, safety pass rate above 98%. Orq runs your suite against every new trace automatically and tracks the scores over time.
Connect Orq to CI and gate releases
Install the Orq CI app or use the status-check API. When a pull request lands, Orq evaluates the incoming branch against your suite. If any metric regresses below its threshold, Orq blocks the merge. No alert fatigue, no post-deploy rollbacks.
Works with every major agent framework
If your agent calls an LLM, Orq can trace it.
Ready to add observability to your agent?
Follow the quickstart and have your first trace in under ten minutes.