Skip to content
ShowcaseShipped

Ten agents.
One product.

Tidy was built by ten agents working in parallel on one product, each change reviewed as a running Worker Preview with its own D1 data.

Agents in parallel
10
Products
1
Review surface
Worker Previews
Preview data
Per-user D1

The problem

Splitting a product across many agents usually ends in merge conflicts and features that pass alone but break together. Reviewers also need to see the app running, with data, not just read a diff.

What Gitflare did

  1. One mission, ten assignments

    Gitflare split the product into ten assignments and started ten agents, each in its own fork and container.

  2. Shared progress

    Agents saw each other's progress, so interfaces lined up instead of colliding at the end.

  3. The whole change, checked

    Gitflare combined the ten revisions into one candidate and ran the checks again on the result.

  4. Worker Previews with per-user D1

    Each candidate ran as a live Worker Preview backed by isolated, per-user D1 databases, so reviewers clicked through real flows before approving.

Cloudflare products

Gitflare features

  • Missions
  • Parallel agents
  • Combined checks
  • Worker Previews
  • Deploy on land

Timeline

  1. Brief

    One product brief, turned into a mission with ten assignments.

  2. Parallel build

    Ten agents in separate forks, sharing progress.

  3. Preview review

    The combined candidate reviewed as a Worker Preview with per-user D1.

  4. Landed

    Approved on the exact revision and landed on main.

Make room
for more minds.

Parallel agents. One reviewed revision.

Get started