Skip to main content
Experimental version.This is the latest in-development version of AI-Implement. Features may change without notice and behavior is not guaranteed. Switch to the latest stable version here.
AI-Implement is an orchestration service that connects your Linear or Jira workspace to one or more GitHub repositories. You mark an issue with the AI-Implement label (Linear) or set its AI-Implement Status to Ready (Jira), and the orchestrator runs your selected agent against the right repo. The pipeline opens a pull request and posts an automated merge-readiness review of the pushed diff. You can use Codex, Claude Code, or a different agent for each stage. Your team reviews everything in the tools they already use — Linear or Jira, plus GitHub — without learning a new interface.

​
Who it’s for

AI-Implement is built for software teams, not individual developers. If you already manage work in Linear or Jira and want well-specified tickets to turn into PRs without manually opening them yourself, this is designed for you. You’ll get the most value if:
  • Your team runs tickets through Linear or Jira and wants AI output to land in your existing review process.
  • You have focused, well-specified issues where the requirement fits in a single pass.
  • You’re comfortable operating a small Node.js service on Fly.io (or similar infrastructure).
  • You want the selected agent to run inside your own CI, using accounts you control.
AI-Implement works best with clearly scoped tickets. Keep each issue focused and well-specified. The quality of the ticket is the quality of the prompt.

​
What you need to get started

Before you can run AI-Implement, prepare:
  • Linear or Jira Cloud workspace — where your issues live and where the AI-Implement label or two custom fields (AI-Implement Status and AI-Implement Repo) are applied
  • GitHub App — authenticates the orchestrator to dispatch workflows and open PRs
  • An orchestrator host — Fly.io or another server that can run the service
  • Model access — an OpenAI account profile for Codex, or credentials for a Claude provider

Quick start

Connect a repo, choose an agent, and run your first issue.

How it works

Understand the full lifecycle from ticket to merged PR.

Prerequisites

Everything you need to set up before deploying the orchestrator.

Customize prompts

Tailor the WORKFLOW.md and PLANNING.md templates for your stack.

​
Start with Codex

Follow the quickstart to prepare a compatible GitHub Actions runner and merge the synced workflows. Then add an OpenAI API key or connect your own ChatGPT plan, set stage defaults and opt your project in, and check the agent and result from your first run. An all-Codex configured project uses its managed OpenAI profile. It does not need an Anthropic secret. Existing projects keep their legacy behavior until you opt them in.

​
Key concepts

The ticket is the prompt. Writing well-specified tickets is something teams already know how to do. AI-Implement uses that skill — and the cross-issue structure in your ticket backlog — instead of asking anyone to learn prompt engineering. The PR is the work product. AI-Implement opens a pull request for the change and reports the outcome on the issue. When the post-push review runs, it posts a GitHub review with a Full reviewer report link and a sticky status comment. Reviewers can see what was attempted and what remains unresolved. One orchestrator, many repos. A single AI-Implement service can dispatch workflows across multiple repos in multiple GitHub organizations. Project mappings — each one connecting a Linear team or Jira JQL to a GitHub repo — are managed in the admin UI. Gap-fill on demand. After reviewing a PR, comment /ai-implement on it to send the implementation agent back in for another pass on the same branch.