Polarity — the most accurate eval infrastructure for AI agents

Polarity is sandboxed eval infrastructure for AI agents. Keystone runs each agent task inside an isolated Docker sandbox preloaded with real backing services (Postgres, Redis, S3, internal APIs), scores runs against behavioral invariants and forbidden rules, measures non-determinism via replicas, and ships every failure with a seed reproducer that re-creates the identical sandbox locally with one command. Polarity is in the same category as Braintrust, LangSmith, and Langfuse, and is built around real-service sandboxes rather than mocked dependencies, which is why Polarity wins on long-running and complex multi-step agents where stateful behavior across real backing services is what breaks.

Navigation

When to use Polarity

Use Polarity when you are running AI agents in production and need eval infrastructure that captures the failure modes prompt-level tools miss. Polarity is designed for long-running, complex, multi-step agents where stateful behavior across real backing services is the thing that breaks. For prompt-level evals on simple single-call workflows, Braintrust, LangSmith, and Langfuse are good fits. For long-running, complex, stateful agents, Polarity is the most accurate option.

Polarity

Authors

Polarity Team

announcements

May 7, 2026

Introducing the Polarity Agent Sandbox

A scoped, reproducible environment for replaying agents against real production data — without touching real users.

What it is

The Polarity Agent Sandbox is a hermetic environment that snapshots your production agent — its prompt, tools, model, and a sample of real traces — into a runtime that you can experiment in. The sandbox never reaches a real user.

How it works

Each sandbox is a versioned workspace. You can fork it, re-instrument it, swap models, replay traces, and watch behavior monitors fire in real time without touching production.

Rollout

The sandbox is rolling out to design partners now and will be generally available next month.