AI Coding Factory
Queue-based workers poll Jira to autonomously implement and push code issues.

Jira-first AI workflow when Cursor and Devin start from chat prompts instead.
Engineering teams using Jira who want to automate small bug fixes and features
Cursor · Devin · Continue
- Pick a Jira ticket - Start an agent - Explain the task - Wait for it to finish - Open a PR - Repeat
So I built FirstDraft.
A worker runs on a machine that already has access to your repositories, build tools, credentials, and anything else it needs. It watches a Jira board, claims a ticket, runs Claude Code or Codex locally, pushes a branch, and opens a draft PR.
The goal was to push through some of those smaller bug tickets and features that could be done without me having to pick them up myself.
What’s been interesting is that once I got the basics working, I started using FirstDraft to build FirstDraft. A lot of the features in the current version were implemented by workers picking up Jira tickets and opening PRs.
It’s still early and very much an experiment, but it’s been a fun one.
Queue-based workers poll Jira to autonomously implement and push code issues.
Full-lifecycle AI dev at $2.5k/mo, but context persistence and code quality TBD post-trial.
Ticket-to-PR automation, but Cursor, GitHub Copilot Workspace, and Devin already own this space.
Ticket-triggered AI engineer when Cursor, Devin, and Sweep already own this category.
Passive standups from Git data, but Geekbot and Standuply already do this.
Sandboxed AI agents coordinate code tasks locally without sending data to cloud.