Empirical CLI
For you and your agents.
Your coding agent changes the app, then asks Empirical for the tests to go with it. One binary, an agent skill, and every API endpoint behind your login.
01 · Install
One binary. Log in as yourself, then install the skill so Claude Code, Codex, and other agents know when and how to call it.
$ curl -fsSL https://cli.empirical.run/install | sh $ empirical login $ empirical skill install --global
02 · Built for agents
Your agent calls the CLI like any other tool. -x waits for
Empirical's agent and prints only its final answer, so it fits in a loop or a script. Add a
CUSTOM.md next to the skill for your project's conventions.
> add coupon codes to checkout · Edited src/checkout/coupon.ts (+84) · Opened PR #418 · Ran empirical session -x "checkout takes coupon codes now (PR #418), add tests for valid, expired, and stacked codes" ✓ PR #1262 opened · 3 tests added · run passed · Coupons are in PR #418; their tests are in PR #1262.
03 · New tests after every change
The work runs in Empirical's sandbox, not on your laptop. It writes or fixes the tests, runs them against your app, and opens a PR an engineer reviews.
Your agent might ask:
- checkout takes coupon codes now (PR #418), add tests for it
- the invite flow changed in PR #412, update its tests
- settings moved under /account, fix the tests that broke
04 · The whole product, from a terminal
Everything in the dashboard is a command away, for you or your agent.
empirical session -x "<prompt>"- Hand work to the Empirical agent and wait for its answer
empirical session logs <id> -f- Follow a session live
empirical clone- Clone your test repo, edit, push a branch
empirical api <endpoint>- Call any API endpoint with your login
empirical skill install --global- Teach your coding agent to use all of the above