Hand the agent the work around your pull requests
Run automations on a schedule or when something happens, and talk to MergeHarbor in Slack, Microsoft Teams, Google Chat or Discord. Code changes go to Shipyard, the coding agent.
Waiting on review in acme/api, posted every weekday at 9:00
3 open pull requests have waited more than a day:
#482 Add a burst option to the rate limiter (2 days) #479 Retry refunds on gateway timeouts (1 day) #471 Remove the legacy export job (1 day)
How it works
Pick what it does
Choose an action from the automations catalog, or ask MergeHarbor a question by mentioning it in a mapped chat channel.
Say when and where
Run on a schedule or when something happens, filtered by repository, team, label and severity, and post to chat, a tracker or email.
Test before you save
Test shows the exact message an automation would post without posting it. Anything that writes code or changes a tracker waits for a person.
What you get
- Automations
- Daily lists of waiting pull requests, risky-merge summaries, incident suspects and more, posted to chat, trackers or email.
- Chat apps
- Ask about a pull request, summarise its review, explain a finding or open a fix pull request behind a confirm button.
- People approve changes
- Anything that writes code or changes a tracker waits for a person with the right permission.
- Code changes in Shipyard
- Shipyard, the coding agent add-on, plans a change, makes it in a sandbox and opens a reviewed pull request.
Frequently asked questions
Do automations use credits?
Every catalog action except Custom instruction works from data MergeHarbor already has and uses no credits. Custom instructions have a per-run cap and a monthly budget.
Can the agent change code?
Code changes are Shipyard, the coding agent add-on. From chat, a mapped channel can hand a task to Shipyard, which posts a plan to approve before any code is written.
Have another question? Ask our support team
Review every pull request with the context of your whole codebase
Connect a code host in minutes. Reviews start right away, and codebase context is added as indexing completes.