Building a custom Jira app used to mean writing Forge code, managing dependencies, and waiting on a developer. No-code tools, like AI Apps Builder, have changed that. Now you can describe what you need, and the AI generates a complete, deployable Forge app for you.
But how does it actually work? This guide walks through every step, from your first prompt to a live app in Jira, and explains what's happening behind the scenes at each stage.
AI Apps Builder for Jira is a secure, no-code platform where AI and humans collaborate to create custom Forge apps. You describe your app idea in plain language, and the system generates a complete Forge app, including modules, permissions, UI, and backend logic.
It supports 31 Jira modules across Jira Core and Jira Service Management, third-party integrations with tools like Slack, Gmail, HubSpot, and Notion, and deploys entirely within Atlassian's Forge platform. Your data never leaves Atlassian's infrastructure.
It's built for Jira admins, product managers, team leads, and solutions consultants who need custom Jira functionality without a development project.
After installing AI Apps Builder, open it from the Apps menu in the Jira sidebar. You'll see a chat interface. This is where everything starts.
Describe the app you want to build. Be specific about what it should do, where it should live in Jira, and what data it should work with. To get more details, read the article: “How to Write Effective AI Prompts: Best Practices and Practical Examples.”
Example prompt:
Generate a Jira Dashboard Gadget called "User Workload". Count open issues grouped by assignee and display them as a simple list or bar chart.
A few things worth knowing at this stage:
AI Apps Builder understands your native language.
Your prompts are never used for training—neither Anthropic (the AI provider) nor the AI Apps Builder team uses your prompts or outputs to train AI models.
You can attach screenshots — instead of describing a design style, upload an image and ask the AI to match it.
If you're not sure where to start, the template catalog includes ready-to-use starting points: OKR boards, Gantt chart planners, meeting agenda pages, subtask creators, and more.
If your app needs to connect to an external service, you can select from the supported integrations list or describe your own.
AI Apps Builder supports three integration types:
Public APIs — no authentication required.
API key — for tools like Notion and HubSpot.
OAuth 2.0 — for Gmail, Google Drive, and Google Calendar.
Supported tools include Slack, Gmail, Google Drive, Google Calendar, HubSpot, and Notion. Any tool with a public or authenticated API can be connected.
When an integration is included, the builder automatically generates an admin settings page inside Jira with setup instructions, credential fields, and resource configuration. You don't build this page; it appears as part of the generated app.
Before any code is generated, AI Apps Builder produces an app specification — a full description of your Jira app. The app specification includes:
Summary
A short overview of what the app does and who it is for.
Modules and Entry Points
A list of Forge modules and where they appear in Jira.
Core Capabilities and User Flows
A description of the app’s main features and how users will work with them.
User Interface Behavior and States
An outline of the UI layout and the main UI states (loading, empty, success, error).
Data and Domain Concepts
Definitions of important terms and concepts used in the app.
Backend Responsibilities and Business Logic
A description of the backend processing, data fetching, and business rules.
External Integrations and APIs
The Jira REST API endpoints used by the app.
Permissions and Scopes
The Jira and Forge scopes are required for the app to work.
Configuration Management
The settings that can be configured and how they are stored.
Out‑of‑Scope Items and Assumptions
What the app does not cover and key assumptions about the Jira instance.
You can:
Review each scope and permission.
Edit the specification if something doesn't match your intent.
Download the spec as a PDF. It is useful for enterprise sign-off or client approval workflows.
This step removes the "black box" problem, because you can see exactly what the app will request access to before it's built. Once you're satisfied, click Approve & Generate App to start the build.
At this stage, wait about 10 minutes while your app gets ready to deploy. Feel free to use this time to take care of other tasks.
What's happening under the hood: AI Apps Builder uses a Large Language Model (LLM) provided through Anthropic. The system operates on an agent-based workflow following a ReAct-style loop — Reason → Act → Review.
It works like this:
The agent “thinks” about which tool it needs to use to complete the app.
AI selects and executes that tool.
AI reviews the result and repeats the loop until the app is ready.
When the app is ready, click Review & Deploy to open the preview environment. This is a live version of your app with real UI, real layout, and real behavior. You can play around with the preview, find any problems, and tweak things with the AI before you deploy. If something doesn’t look right, just tell the AI what to change in the chat, and it will update the app. Then you can check the preview again. This is the stage between "the AI built it" and "my team sees it."
AI Apps Builder offers two deployment options: Manual and Automatic.
Download a ZIP file of the generated app, run the Forge installer for your operating system, and enter three things: your Atlassian account email, your Jira domain, and an API token.
Fill in your Jira domain and API token directly on the deployment page, click Test Connection, then click Deploy to Jira Cloud.
Both methods require the same API token. Two things worth knowing upfront:
The token is used only for that authentication step
AI Apps Builder does not store your API tokens
Once deployment is complete, the built app runs independently as a standard Forge app. If the token expires or is revoked later, the deployed app continues to function normally. The token doesn't play any role after installation.
To create an API token: Go to Atlassian Account Settings → Security → Create API token → label it "AI Apps Builder" → copy and paste into the deployment field.
After deployment, the page shows your app name, placement (where to find it in Jira), deployment date, and status.
You have three paths to improve your app:
Open the AI Apps Builder sidebar, select your app, choose the version you want to refine, and describe the changes you need. After the AI updates the app, check the preview and redeploy. Version history saves every state automatically, and you can roll back to any earlier version at any point.
Download the ZIP file of the generated source code from the deployment page. Hand it to a developer who can extend or customize the Forge app directly. You receive a working, deployable app to start, and the developer continues from there.
If the AI-generated output has issues you can't resolve through prompting, the Dev Help feature connects you with real developers who will:
Review the generated code
Fix generation issues
Improve functionality
Add custom logic based on your requirements
Send us a request explaining what's not working, what you expect the app to do, and include any helpful screenshots or files. Our team will create a support ticket, check out your app, and work with you until it meets your needs.
Yes, you can. Describe what you want in plain language, and AI Apps Builder handles the rest:
AI Apps Builder generates complete Forge apps from plain-language descriptions — no code required.
An app specification is generated before code, so you review permissions and scope before approving the build.
Third-party integrations with Slack, Gmail, Google Drive, Google Calendar, HubSpot, and Notion are supported — with admin configuration pages generated automatically.
Live preview lets you verify UI, layout, and behavior before deployment.
Version history saves every state — roll back, branch, or experiment freely.
The entire Forge app — modules, permissions, UI, authentication, and backend logic — is generated automatically. You see a simple chat. The complexity stays out of your way.
Your data stays protected throughout:
Your prompts are never used to train AI models.
All Jira data stays within Atlassian's infrastructure.
Install AI Apps Builder for Jira and start with something your team already needs — a custom dashboard, an issue panel that pulls data from an external tool, or a JSM portal component that matches your brand.
You start with 100 free credits. Enough to build something real and see how the whole process works from prompt to deployed app.
Mariia_Domska_SaaSJet
0 comments