Use cases

Oqoqo fits wherever agents do real work on your product. These are the most common ways teams use it. They share the same loop: define a task, run it across agents and treatments, and improve from the evidence.

​

Test agent-facing interfaces

Your SDK, API, CLI, MCP server, and skills are how agents use your product. Run real agents against those interfaces, see where they slow down or fail, and fix the interface before your users’ agents hit the same wall. Use treatments to compare an interface against the raw agent.

​

Run agent evals at scale

Stop testing agents by hand. Define a workflow once as a task with a rubric, then run it across agents, models, and treatments and get consistent, comparable pass rates every time. See Tasks and rubrics.

​

Build private benchmarks

Turn your real user journeys into a benchmark suite you own. Because tasks and rubrics are versioned, you can run the same suite against any agent version and any product version, and show in numbers whether you are getting better.

​

Gate releases in CI/CD

Block changes that break agent workflows, not just unit tests. Run an experiment as part of your pipeline, and fail the build when the pass rate drops below your bar.

​

Catch model and dependency drift

Models and dependencies change under you. Rerun the same experiment whenever a model updates or a dependency bumps, and compare against your last known-good pass rate to catch a regression before your users do.

​

Optimize model spend

The most capable model is not always needed. Run the same tasks across models, compare pass rate against cost and tokens in one view, and move to a cheaper model where it still passes.