Automatic, on every PR
Install the GitHub App and connect the repo. Kery picks up each pull request with its preview deployment, runs the flows the diff touches, and posts the verdict. No workflow YAML, no runner.
Composer and background agents open pull requests faster than anyone can open the app and click through them. Kery closes that gap without adding a suite to maintain.
The whole point of a background agent is that you were doing something else. That also means nobody sat with the change, nobody opened the app, and the first human contact with the work is a code review of a diff.
Reading a diff tells you the code is plausible. It does not tell you the sign-in still works, the form still submits, or the mobile layout did not collapse.
Install the GitHub App and connect the repo. Kery picks up each pull request with its preview deployment, runs the flows the diff touches, and posts the verdict. No workflow YAML, no runner.
Add the Kery MCP server to .cursor/mcp.json. Ask Cursor to run a check on a
flow in plain English and read the result without leaving the conversation.
Kery ships an MCP server, which Cursor picks up from your project's .cursor/mcp.json or from Cursor's MCP settings. Once it is connected, you can ask Cursor to run a Kery check on a flow and read the result back in the same conversation.
No. In-editor browser access is useful while you build. Kery is the pass that runs on the pull request afterwards, without being asked, and posts a verdict the whole team can see.
Yes, and it is a good fit. Background agents open pull requests you were not watching, which is exactly when an automatic browser check earns its keep.
Connect the GitHub App and point Kery at your environment and credentials once. After that, each pull request is picked up along with its preview deployment.
Running the open-source engine locally, yes — bring your own OpenAI, Anthropic, Gemini, or OpenRouter key. Kery Cloud handles execution for you.
In the open-source setup, in your own environment. Kery only sends prompts and the context a run needs to the provider you configured.
Connect a repo and Kery starts checking pull requests against their preview deploys. No test scripts, no CI config.