Loop Recipes
Ready-made multi-agent teams. Each recipe spins up a whole team of AI agents — with models, repos, toolpacks, skills, and first prompts already wired up. Open one to use it on chat.dev.
BugFinder
Creates a team of agents that act like users and finds bugs. Emails you every few hours with what its found.
Growth Squad
A marketing team that ships campaigns: a strategist researches the audience, a copywriter drafts the assets, and a publisher commits them to your marketing-site repo and sends the announcement email.
Deep Research Pod
Fan-out researchers plus a synthesizer produce a cited report on any question, using real web search and scraping APIs. Findings are committed to a Markdown report in your repo.
PR Review Crew
Two agents that triage your open pull requests on a GitHub repo: one reviews correctness, security, and house style; the other checks out each branch and runs the full test suite.
Content Studio
Turns a brief into a published blog post on your Git-backed site: a writer drafts, an editor sharpens for clarity and voice, and a publisher commits the Markdown and opens the PR.
On-call SRE
An incident team: a monitor watches Sentry/PagerDuty for alerts and opens an incident, a diagnoser investigates the repo and recent changes for root cause, and a fixer ships the mitigation PR.
Sales Outreach
Builds and runs a personalized outbound sequence: a prospector pulls and researches accounts via Apollo, a writer drafts tailored emails, and a CRM agent sends them via Resend and logs activity to HubSpot.
Full-Stack Feature Team
Ships features end-to-end against a shared GitHub repo: an architect writes the API contract and delegates, backend and frontend build in parallel on branches, and a reviewer gates every PR before merge.
Bug Bash Squad
Turns a bug report into a merged fix: a reproducer writes a failing test from Sentry/issue details, a fixer makes it pass, and a verifier confirms no regressions before the PR goes up.
Data Analytics Pod
Answers a data question end-to-end against your database: an extractor pulls and validates the data, an analyst explores and models it, and a reporter commits a written, charted findings doc to the repo.