Each recipe is a working defineScore snippet using real, published @tuttiai voices. Copy the score, install the listed voices with tutti-ai add, drop your provider key into .env, and run.
Fetch a GitHub PR, read the affected files from the working tree, and post a structured review.
Connect to Postgres, introspect the schema, and answer business questions without writes.
Look up customers and charges, surface details, and propose refunds — but never issue them without operator approval.
Ingest your docs into a vector store and answer with citations. Says "I don't know" when retrieval is empty.
Search, fetch, summarise — with citation metadata returned by the Web voice on every result.
Read open issues, code-search the repo for context, and propose labels + assignees — never closes without approval.
Read messages flagged in #reports, classify each, and propose moderation actions for an operator to approve.
Open a PR or an issue with your score and a short writeup of what it solves. We curate the gallery, not gatekeep it — accepted recipes get a permanent slug and an authored byline.
One install. Your first agent running in 60 seconds. No signup. No telemetry.