AI Developer
Integration of Claude Code / OpenAI Codex
AI Developer is your autonomous AI engineer powered by Claude Code / OpenAI Codex inside Jira. It takes over issues directly from your backlog, analyzes the task, and delivers working results within minutes. Whether it'prototyping new features, resolving bugs, or accelerating delivery, AI Developer transforms ideas into reality without disrupting your existing workflow.
With AI Developer, your team gains the ability to prototype faster, code smarter, and remove bottlenecks from the development process.
When assigned to an issue, AI Developer generates a solution, pushes the implementation to a dedicated Git branch and shares the outcome as a comment in Jira - ensuring transparency and keeping the team in control. Each assignment runs in a secure, isolated environment to protect your code and maintain full data integrity.
AI Developer
An autonomous AI engineer, powered by Claude Code / OpenAI Codex, that runs inside Jira. Assign any issue and it will analyze the task, generate a solution, and deliver working code — typically within minutes.
How it works
What AI Developer does
- Reads the issue summary and description
- Generates a solution — code, tests, and dev notes as applicable
- Pushes the implementation to a dedicated Git branch
- Posts a comment on the Jira issue with the outcome
Works best for clearly scoped tasks: prototyping features, resolving bugs, or accelerating delivery without disrupting your workflow.
Features
Added to every Jira issue page. One click hands the issue to the AI.
Generates a complete proposed solution — code, tests, and dev notes — directly from the issue summary and description.
Commits changes and pushes to a dedicated branch in your configured repository.
A comment is created on the issue with the solution summary and branch name for full transparency.
Configuration
You need three things before AI Developer can push code: an HTTPS Git connection, a repository access token, and a Claude/Codex API key.
App setup
- Open Project / Space Settings → Apps → AI Developer
- Enter your HTTPS repository URL, e.g.
https://github.com/your-org/repo.git - Enter the base branch to start from, e.g.
main - Enter your Git username and access token as the password (see token guides below)
- Optionally set an identifier to differentiate multiple repositories
- Choose your AI connection method and enter the relevant credentials (see below)
AI connection
You can use AI directly from the AI Provider Anthropic/OpenAI or use it via AWS Bedrock or Microsoft Foundry (only supported by now by the "Integration of Claude Code"; availability for "Integration of OpenAI Codex" with the next release).
Create a Claude API key / Codex API Keyand paste it into the LLM API Key field.
Create a long-term (or short-term) API key in the Amazon Bedrock console and select your AWS region.
One-time setup required for Claude: Submit the Anthropic model use case form in the Model catalog before first use. Go to the Amazon Bedrock console, open the Model catalog, select an Anthropic model, and submit the form — access is granted immediately.
Provision a Claude/Codex model in the Microsoft Azure AI Foundry portal first. Then enter the resource name and API key, and make sure the model selection in the settings matches your provisioned model.
Git access tokens
Use a Personal Access Token (PAT). Fine-grained tokens are recommended for tightest control.
- Go to Settings → Developer settings → Personal access tokens → Fine-grained tokens
- Click Generate new token, give it a name, and set an expiry date
- Under Repository access, select only the repositories AI Developer should act on
- Set permissions: Contents read & write · Commit statuses read & write · Pull requests read & write (if you want PRs created)
- Generate the token, copy it, and paste it as the HTTPS password in Jira settings
Use a Personal Access Token or a Project Access Token (scoped to a single repo).
- User PAT: Profile → Preferences → Access Tokens. Project token: Project → Settings → Access Tokens
- Give it a name, optional expiry, and select scopes:
read_repository,write_repository(addapiif merge request creation is needed) - Create the token, copy it, and paste it as the HTTPS password in Jira settings
Use an App Password (or API token on newer Bitbucket instances).
- Go to your avatar → Personal settings → Access management → App passwords
- Click Create app password, give it a label, and select scopes: Repository read & write · Pull requests read (and write if you want PRs)
- Create the token, copy it, and paste it as the HTTPS password in Jira settings
Usage
After installation, an Assign to AI Developer button appears in the context panel of every issue. Click it to send the issue to the AI for resolution.
Writing effective issues
AI Developer has no project memory — everything it needs regarding requirements must be in the issue description. Well-written issues produce significantly better results.
- Business requirements — goals, acceptance criteria, inputs/outputs, constraints
- Technical context relevant to this specific task
- Expected vs. current behavior (for bugs)
- Use bullet points or numbered lists for clarity
- Separate technical details from business requirements
- Break large tasks into multiple smaller issues
If the first result isn't right, refine the issue description — add more context, rephrase, or restructure — and re-assign.
Support
For help or to report issues with the plugin, email support@smart-fusion.tech.
Security tips
Prefer machine or service accounts for tokens rather than personal user accounts.
Always set an expiration date when creating tokens and rotate them periodically.
Restrict tokens to the minimal set of repositories and scopes required. Use fine-grained tokens where supported.