Checksum AI

Checksum is a continuous-testing platform aimed at engineering teams whose coding agents now ship changes faster than a human QA function can cover them. Rather than describing a test case by hand, you hand the platform an objective — "reach 90% coverage on the Activities feature" — and an autonomous agent explores the product, works out which flows matter, and writes the tests.

Three agents, three jobs

• E2E Agent — bootstraps broad coverage from a cold start, typically 100–150 tests in the first week.
• CI Agent — runs on every pull request and generates 50–200 tests aimed specifically at the code that changed.
• API Agent — extends coverage across endpoints, including multi-endpoint verification that a single request-level assertion would miss.

Maintenance is the real product

What separates Checksum from record-and-replay tooling is what happens after the tests exist. When a selector moves or a flow is redesigned, the auto-healing agent detects the failure, repairs the affected test and opens a pull request for review — the company reports roughly 70% of broken tests resolving without human intervention. The agents work autonomously against a stated goal instead of requiring turn-by-turn prompting.

No lock-in

Everything Checksum produces is ordinary Playwright code committed to your own repository, and it runs inside your existing CI on every commit, pull request and deployment. If you stop paying, the suite is still yours and still runs.

Pricing and adoption

Checksum bills on a "Results-as-a-Service" basis tied to the number of workflows it maintains, rather than per seat or per test run; figures are quoted by the company rather than published, so expect a sales conversation. It lists 30+ customers, among them ActiveCampaign and LiveRamp.