New · AI Developer

Spec Driven Development in Jira

Agree on what to build before the AI writes a line of code. A Jira ticket becomes a spec package that your team reviews and edits right in Jira.

Then the AI applies it, verifies the result against the spec and archives the change, with everything stored in your git repository.

See it in action

Loads the video from YouTube when you click play. Watch on YouTube

Why spec-first?

Review the plan, not just the code

Vague tickets lead to wrong code that is expensive to review. A spec gives your team a cheap checkpoint before implementation: what changes, why, how, and which tasks.

Humans stay in control

Every artifact is editable in Jira and in Git. Edit it yourself, or describe the change in plain language with Change spec. Your edits are respected when the AI applies the spec.

Built-in verification

After applying, the AI checks completeness, correctness and coherence against the spec and reports findings as CRITICAL, WARNING or SUGGESTION.

How it works

Create spec

The AI reads the ticket and your repository and writes the proposal, specs, design and tasks, then commits them to an ai/* branch.

Review and edit in JiraOptional

Each artifact opens as a tab in the issue panel. Edit it in the rich-text editor, or ask for changes with Change spec.

Apply spec

The AI implements the open tasks, ticks them off and merges the spec changes into your living specs, on the same branch.

Verify changes

A report checks completeness, correctness and coherence and ends with a clear verdict. It is posted as a Jira comment.

Finish spec

The change is archived in git, and the issue is done. Start a new spec for the next change.

What you get

A complete spec package

Proposal (what and why), one spec per capability, design (how) and a task checklist. The format comes from OpenSpec and is stored as plain Markdown in your repo.

Everything in git

Spec and code live on the same ai/* branch. Applying a spec syncs it into your main specs under openspec/specs, so living documentation stays current as a by-product.

Your AI, your infrastructure

Choose Claude or OpenAI Codex per Jira project. Connect with an API key, a subscription, AWS Bedrock or Microsoft Foundry. Jobs run in isolated runners with regional options.

Frequently asked questions

Do I need to install OpenSpec?

No. OpenSpec is built into the runner. It does help to initialize OpenSpec in your repository (openspec init) so the folder structure is in place, but it is not required.

Can I edit the spec?

Yes. Every artifact can be edited in Jira until you apply the spec. You can also describe a change in plain language with Change spec, and the AI updates only the affected artifacts while keeping your edits.

Which AI providers can I use?

Claude or OpenAI Codex, selected per Jira project. Connect with an API key, a Claude or Codex subscription, AWS Bedrock or Microsoft Foundry.

Does it work with existing repositories?

Yes. The AI reads your repository to write a spec that fits your code. A CLAUDE.md or AGENTS.md in the repo root improves the results.

What does it cost?

Pricing and licensing are the same model as the Claude Code and Codex apps and are shown on the Atlassian Marketplace page. AI usage is billed by your own AI provider.

Get Spec Driven Development

Available on the Atlassian Marketplace.

Get Spec Driven Development Read the docs