Ask in Slack. It answers in Slack. And it remembers.
You already tried this in Slack. The thread did not survive.
Agents
CI run 1 · CI run 2 · same thread
Hatch keeps one conversation per pull request, however many times CI fires. Anyone on the team can open it.
Illustrative product example. No live agent.
CI fired again. The reviewer that already left comments did not have them. The session link opened on one person's account. Nobody else could pick the work up.
Hatch keeps one conversation per pull request, however many times CI fires. Anyone on the team can open it.
Each colleague has its own computer. The repo stays cloned. The notes it wrote itself are still there on the next turn.
You talk to it in Slack, on a pull request, or in Hatch. It answers in that same place.
Label the PR. Reviewer comments on the PR as itself, in the same thread as last time. Mention it in Slack. It replies in that thread.
You ask Hatch. It routes the question, or it makes you a colleague. The file it writes is there afterwards, to read, review, and edit.
You do not start in YAML.
Two defects. Nobody planted them.
Engineering teams of about 5 to 50, on GitHub and Slack. You have already tried an agent in the thread.
Request access. It starts a Hatch of your own.