AI Apps Builder in 2026: What Changed, and Why It Matters

[#1]

The start of 2026 has been a big turning point for AI Apps Builder for Jira. If you’ve been following the product for a while, you already know that a lot has changed—not just visually, but in how the tool actually works and how teams use it day to day.

This overview walks through what changed, why it matters for Jira users, and where we’re heading next.

From No‑Code Apps Creator to AI Apps Builder

The first visible change was the name. We renamed No‑Code Apps Creator to AI Apps Builder. The old name focused on how the app was built (no code). The new name reflects what actually creates the value: AI that actively helps you build better custom Jira apps on Atlassian Forge.

Today, AI Apps Builder is a builder that:

• Understands user intent from natural‑language prompts

• Makes decisions about which tools and steps to use

• Fixes mistakes during code generation

• Improves results over time as the agent learns from more cases

That shift set the direction for everything else we shipped this year.

The biggest change: an agent model that thinks and fixes

The most important improvement at the beginning of 2026 is the move to an agent model. AI Apps Builder no longer “generates once and hopes it works”. Instead, it follows an agent‑style ReAct loop: plan → take an action (generate or fix) → review → repeat until the app is ready.

That’s why you can now get a usable Forge app from a single, well‑written prompt, instead of manually tweaking code or constantly regenerating. For Jira non‑technical users, this turns “ask an LLM” into “ship a working app”.

Previously, AI Apps Builder worked more like this:

1. You enter a request to generate an application

2. The LLM generates structured output

3. A response parser splits that output into separate files

Now it works more like this:

1. You enter a request to generate an application

2. The agent “thinks” about what it needs to do to complete the task

3. AI decides whether it needs to call a tool

4. AI selects and executes that tool

5. AI reviews the result and repeats the loop until the app is ready

From your point of view, nothing got more complicated: you still write one prompt—AI does the thinking work behind the scenes.

A clearer, more user-friendly UI

Another big change this year is the updated UI. We didn’t redesign the interface just to make it prettier. The goal was clarity.

We focused on:

• Adding more color and structure so it’s easier to orient yourself.

• Simplifying flows for generating and updating Forge apps.

• Making “next steps” obvious after each action.

• Bringing security information into the main experience instead of hiding it in docs​​.

Making security visible

Often, Jira teams hesitate to adopt AI tools. Security and compliance teams need clear boundaries around where Jira data goes and who can access it.

Security has always been part of AI Apps Builder because everything runs on Atlassian Forge. In 2026, we made that security model clear and understandable:

• Allowed users to review all requested scopes before deployment, so users know exactly which data the app can access.

• Added a modal with brief security information.

Data security

• The app is built and deployed on Atlassian Forge, Atlassian’s secure application platform.

• The app runs entirely within Atlassian Cloud.

• The app follows Atlassian’s security, compliance, and access‑control standards.

• Jira data is processed and stored only within the Atlassian infrastructure.

• Data access is governed by standard Jira permissions and approved scopes.

• No data is sent to external servers, third‑party services, or external AI systems.

• Site administrators have full transparency and control over data access.

AI security

A key point for many teams: the AI does not see your Jira data.

• The AI does not access, read, or analyze any Jira issues or fields.

• It never connects to Jira or uses Jira APIs directly.

• AI is only used to generate Forge app code based on your prompts and public documentation​.

Jira data is only accessed after deployment, inside Atlassian Cloud, and always under Forge permissions and Jira’s existing security model.

Credits: making AI usage predictable

We also changed how we track LLM effort and moved to a credit‑based model, similar to other modern AI and agent platforms.

AI Apps Builder is a free Jira app, and to make exploration easier, every new user gets 100 free credits. Credits are used when you send messages in the AI chat and ask the agent to generate or update an app.

Each message uses a certain number of credits based on:

• The length of your message.

• The complexity of the requested functionality.

• The number of reasoning steps the AI performs.

• The effort required to generate or update a Forge app, including UI, logic, and configuration.

Now users can:

• See how many credits they’ve used.

• Understand how many credits a build or improvement costs.

• Estimate what they need before starting a bigger idea.

Example: how credits are used in AI Apps Builder

This example shows how a single prompt that generates a custom Forge app with UI, logic, and data handling consumes credits based on complexity.

Initial app generation

Prompt:

“Create an Issue Panel for Jira that shows an overview of the current sprint.
The panel should include:
– a timeline view showing the total number of issues in the current sprint and overall sprint progress
– a list of issues where the due date has already passed (highlighted in red)
– and issues that are due within the next two days.”

Used: 4.6 credits

Small edit

Prompt:

“Add the ability to open an issue when clicking on it.”

Used: 1.6 credits

Simple requests use fewer credits. More complex app generation or advanced changes consume more, because the agent needs more steps to reason, generate, and verify the result.

This is just the beginning for AI Apps Builder

What is different in 2026 is not just a new name.

• The foundation is stronger.

• The product is clearer for Jira users.

• The AI agent is smarter and more resilient.

• Creating loops from idea → prompt → working custom app are better.

Most importantly, we’re building this with users, not just for them. Every improvement this year came from:

• Real Jira use cases.

• Real friction in real projects.

• Real feedback from admins, PMs, and teams who live in Jira every day.

AI Apps Builder in 2026 is not a finished product—it’s a new phase. We’re working every day to:

• Improve the agent model and its internal tools.

• Make results more predictable and repeatable.

• Reduce effort for non‑technical users.

• Help teams shape Jira around how they actually work, not the other way around.

If you’ve tried the product before, now is a good time to look again. If you’re new, you can start small, experiment with free credits, and see what’s possible in your own Jira site.

And if you have feedback—that’s exactly how the next version of AI Apps Builder gets built.