These agents do not wait to be asked. They watch the work where it happens — pull requests, issues, calls, release feeds — and take the next step on their own.
Agents that live in your repo and move pull requests from opened to merged.
When a PR opens or gets new commits, run a senior review pass and post findings so the author never waits on a manual first review.
After finding issues, push fixes straight to the PR branch instead of just leaving comments — including cleaning up other bots’ comments.
Inspect the checkout when checks fail or comments arrive, resolve failures and conflicts where possible, and flag when a human is actually needed.
Send a threaded Slack update when the PR is ready, and merge automatically once an authorized approver signs off.
A hygiene reviewer that keeps quality from sliding as the codebase grows.
Diagnose PRs for duplicated implementation paths, dead code, stale docs/skills/rules, and concrete code smells before the change lands.
Create a Notion journal page per run so maintainers can see what was flagged, what repeated, and where divergence is accumulating.
Recall prior hygiene runs for the same repo and use workspace memory to spot recurring divergence instead of treating each PR as isolated.
Agents that take a request — in Linear or on a call — and open the PR that does it.
When a matching Linear issue is created, classify it and start a coding workflow that opens a GitHub PR and comments the link back to Linear.
When mentioned in a Linear comment or agent session, keep thread memory, decide between discussion and implementation, then reply or delegate coding.
When a Granola note arrives, detect a prospect feature ask, create the Linear issue, implement the change, open the PR, and link it back.
Monitoring agents that watch the outside world and bring only the signal to you.
Twice a day, scan HN titles for configured keywords, summarize fresh matches, and post a Slack digest — deduped so nothing reposts.
Each weekday morning, check configured npm packages against remembered versions and post only the changed packages to Slack.
Daily, check followed Spotify artists for new albums or singles and DM you the release links newer than the last check.
Split a single issue across specialized agents that implement and review in parallel.
A team lead can launch a roster where one agent implements the issue in a sandbox and another audits the branch against the spec before handoff.
The implementer makes the smallest complete change and opens one PR; the reviewer verifies tests and classifies actionable findings.
Browse the full gallery and launch any of these agents on Agent Relay in one click.
Explore all agents →