The TestSprite CLI is now live — open source.Star it on GitHub

The Testing Tool Codex Can Drive Itself

Codex ships code quickly, and reviewing what it produced is now the slow part. TestSprite closes the loop: one command installs a verification skill into your repo, and from then on Codex can create tests, run them against a real browser and live APIs in the cloud, and pull back a single self-consistent failure bundle when something breaks — without you relaying instructions from a docs page.
Type
Solution
Language
English

Seamlessly Integrates With Your Favorite AI-Powered Editors

Claude CodeCodexVisual Studio CodeCursorTrae
We don't write your code — Codex does that. We verify what it wrote, against a real browser and real APIs, and hand back a root-cause hypothesis it can act on.

One Command to Onboard

testsprite setup --agent codex configures your API key, verifies it, and drops a verification skill into your repository. Codex learns the loop once instead of re-deriving it from documentation every session. Installation is purely local.

Tests Without Browser Code

A test is a plain-language plan file with action and assertion steps — no selectors to maintain, nothing that breaks on the next redesign. testsprite test create --plan-template prints a schema-correct skeleton pinned to your installed version.

One Failure Bundle, Not Five Artifacts

test failure get returns the failing step, its neighbours, screenshots, DOM snapshots, the test source, a root-cause hypothesis, and a recommended fix target — all sharing one snapshot id. The CLI refuses to stitch data from two different runs.

A Suite That Outlives the Context Window

Every passing test is banked. The requirement keeps being checked next session, next refactor, and after Codex has long forgotten why it mattered — which no context window, however large, can do.

Priority
Test
Status
HIGH
TC001_Checkout_Submits_With_Valid_Card
Pass
HIGH
TC002_Session_Survives_Token_Refresh
Failed
MEDIUM
TC003_Search_Returns_Ranked_Results
Pass
LOW
TC004_Empty_Cart_Shows_Guidance
Pass
MEDIUM
TC005_API_Rejects_Malformed_Payload
Warning

Let Codex Verify Its Own Work

Generation is fast; review is the bottleneck. Give Codex a way to prove a change works before it opens a pull request, and the diff stops being something a human has to read line by line.

Built for Teams Shipping Codex Code

Free to Install, Open Source

The CLI is Apache-2.0 on GitHub and free to install from npm. Requires Node 20.19+, 22.13+, or 24+. Test execution runs in the cloud and consumes workspace credits.

Works Non-Interactively

TESTSPRITE_API_KEY=sk-... testsprite setup --from-env --yes --agent codex never prompts, which is what an agent loop and a CI job both need.

Safe to Explore

--dry-run exercises the full code path offline with canned data. test scaffold and test lint never touch the network or your credentials at all.

Scriptable by Contract

A stable --output json shape and documented exit codes: 0 passed, 1 failed, 3 auth, 12 out of credits. A broken feature and a broken pipeline are distinguishable.

Trusted By Businesses Worldwide

"Good job! Pretty cool MCP from TestSprite team! AI coding + AI testing helps you build better software easily!"

"TestSprite offers rich test case generation, clear structure, and easy-to-read code. It also supports simple online debugging with the ability to quickly expand by generating new test cases."

"TestSprite's automation helps us reduce tons of manual work. The developers can easily catch and resolve bugs earlier in the development process."

FAQ

How does TestSprite work with Codex?

Run testsprite setup --agent codex once. It installs a verification skill file into your repository describing the commands, the JSON shapes, and the exit codes, so Codex can create, run, and triage tests on its own. Codex is one of eight harnesses the CLI installs skills for, alongside Claude Code, Cursor, Cline, Antigravity, Kiro, Windsurf, and Copilot.

Does Codex have to write browser automation code?

No. Tests are plain-language plan files with action and assertion steps. Run testsprite test create --plan-template for a schema-correct skeleton, or ask TestSprite to propose the first set with testsprite test plan generate — proposals are staged for review and nothing is written to disk until you accept them.

What exactly does TestSprite test?

Web application UI in a real cloud browser, and backend APIs. It does not drive native iOS or Android UI, desktop applications, or emulators. If your product has a web frontend or an HTTP API, that is what we cover.

Is it free, and is it open source?

The open-source CLI is free to install from npm and Apache-2.0 licensed. Running tests executes in TestSprite's cloud and consumes workspace credits — 0.5 per frontend run, 0.2 per backend run. There is a free community tier to start on.

Why not just let Codex run unit tests?

Unit tests confirm the code does what it was written to do. They never open the deployed application, so a green unit suite sits comfortably alongside a broken sign-up flow. Both are worth running — they answer different questions.

Can I run it inside a CI job rather than interactively?

Yes — that is the normal case. The CLI needs only TESTSPRITE_API_KEY in the environment, no credentials file, so it drops into GitHub Actions, GitLab, CircleCI, Jenkins, or Azure Pipelines. Add --report junit for a sidecar those systems ingest natively.

Give Codex a Way to Check Its Own Work