Kaloko for product teams
Several agents work on several branches, and it all ends up in one release. Kaloko gives the product owner, QA and design one place to see what each change really looks like, against the task, and to accept or return it before the merge.
Sounds familiar
- The pull request says “done”; nobody opened the branch.
- The release candidate differs from the accepted design and nobody can say where.
- Three agents worked on one flow and nobody remembers what changed.
How Kaloko helps
- A link in every pull request
kaloko share --prkeeps one comment per scenario; the GitHub check turns green when the run is accepted. - Regression before a release
Walk the flows on staging and compare with the accepted run: statuses, criteria, pixels.
- Design as the reference
An accepted design version becomes the reference; the design pack checks each implementation against it.
One prompt to start
Read the task in docs/tasks/TASK-123, write its acceptance plan and a Kaloko scenario, walk it on the review environment and share the result on the pull request.What comes out
LIVE · today 14:44 CEST · 7 of 11 passed · 7/11 steps · Open the live run ↗
FailedHomepageCS · desktop
OKHomepage na kaloko.appCS · desktop
OKZaložení organizaceCS · desktop
OKPřihlášeníCS · desktop
Waiting for a person· 2 Oct 2026· 36 screens· CS, EN
- OK passed or approved by a person
- Returned a person sent it back with a note
- Failed a check failed
- ? a person decides
Situations
Which plan fits?
Team adds required approvers, MCP and hosted AI evaluation (1,000 questions a month); Business keeps accepted runs forever. Compare plans →
Questions
- Does it replace our end-to-end tests?
- No. A test suite guards yesterday’s flows against regressions; Kaloko accepts each new change against its task and shows it to the people who decide. Many teams keep both: the suite in CI, Kaloko on the pull request.
- Which coding agents does it work with?
- Claude Code, Codex and Cursor through the same skill, any other agent through the CLI, and agents that speak MCP through the MCP server (Team and up). A person can walk a scenario too.
- What does the product owner need to install?
- Nothing. They open the link from the pull request in a browser, sign in with a one-time e-mail link and accept the run or return it with a note. Viewers are free in every plan.
- Where does the walk run?
- Wherever the CLI runs: on a developer’s machine or headless in CI. It walks any address you give it, from localhost to a staging environment behind a gate, and uploads the run to the canvas.
Start with one task
Free plan, free for good · e-mail sign-in · the first run in ten minutes