Connect Claude to your Atlassian — safely

What does this app do?

AI assistants like Claude can now read and change things in Jira and Confluence for you. That's powerful — but it raises two questions: who decides what the AI is allowed to touch, and how do you know what it actually did?

AI Governance sits in the middle. Claude connects to AI Governance instead of connecting to Atlassian directly. Every request the AI makes passes through two checks before anything happens:

• Is this allowed? — Your rules decide what each AI, person, or group may do.
• Write it down. — Every action is recorded: who did it, what they did, and when.

Three reasons teams use it

• Control — Allow or block actions by AI, by user group, or by type of action — for example, “never let any AI delete anything.”
• Visibility — A complete, searchable record of every AI action — including the exact request and the response size, details Atlassian's own audit log doesn't show.
• Safety — Stop risky actions before they happen, not after — and sign in as yourself so every action is tied to a real person.

Part 1 · Connect the app to Claude

Goal: by the end of Part 1, Claude is talking to your Atlassian through AI Governance.

1: Install AI Governance from the Atlassian Marketplace

• Open the listing on the Atlassian Marketplace and click Get it now (or Install). Pick your Jira site and confirm. Installation takes under a minute.
• Search the Atlassian Marketplace for “AI Governance by miniOrange.”
• Once it's installed, open it inside Jira: top menu Apps → AI Governance. (On some sites it appears under Settings → Apps.) You'll land on the Getting Started page.

    Find it on the Atlassian Marketplace

2: Allow this domain in Rovo (one-time setup)

This one-time step makes sure every AI tool is forced to go through your governed URL — instead of quietly connecting straight to Atlassian, where nothing is checked or logged. You'll find this card on the Getting Started page, just below your MCP server URL. (Rovo is Atlassian's own AI service that your app connects through.)

• 1 Callback URL + Copy. The address to add to Rovo's allowlist — it's your site's MCP server URL with /callback on the end. Click Copy.
• 2 The four steps. Do them in order on the Rovo side (spelled out below). When you're done, the card's status flips from Action needed to Allowed.

What to do, step by step

1. Copy the callback URL shown on the card.
2. Open your Rovo MCP server → Domains → Your domains, and click Add domain.
3. Paste the URL and save. The card's status turns Allowed (green).
4. Turn off “Allow Atlassian supported domains.” This is the important one — it forces every AI tool to use your governed endpoint, so nothing can reach Atlassian through an unlogged, native connection.

3: Copy your MCP server URL

This is your site's private address for AI assistants. Claude connects to this URL instead of connecting to Atlassian directly — and that's exactly what makes every call pass through your rules and your log.

• 1 Your MCP server URL. A unique web address for your site. It looks like https://…hello.atlassian-dev.net/x1/… — you don't type it, it's generated for you.
• 2 Copy. Copies the URL to your clipboard. Keep it handy — you'll paste it into Claude next.

4: Paste the URL into Claude

Now hand that URL to Claude:

1. Open Claude → Settings → Connectors.
2. Click Add custom connector.
3. Paste the URL you copied into the URL (or Server URL) field.
4. Leave any “advanced” OAuth fields blank — Claude fills those in automatically.
5. Click Add / Connect.

5: Sign in once — and you're connected

Claude will open an Atlassian sign-in. Approve it. This proves who you are, so AI Governance can apply your rules to you and record every action under your name. You only do this once per connection.

Part 2 · Set the rules & watch the activity

Claude is connected. Now decide what it's allowed to do — and keep an eye on what it does.

2A: Set your access rules — the Access Policy

Open the Access Policy tab in the left menu.

A policy is a default decision plus an ordered list of rules. For each AI request, AI Governance reads your rules from top to bottom and uses the first one that matches. If none match, the default decision applies.

• 1 Default decision. The fallback used when no rule below matches. Allow = let things through unless a rule blocks them. Deny = block everything unless a rule allows it.
• 2 Rules. Each row is one rule, read left to right (explained in the table below).
• 3 Add rule. Adds a new row at the bottom. Order matters — the first matching rule wins, so put your most important rules near the top.
• 4 Save policy. Applies your changes. It takes effect on the very next AI request.

What each column in a rule means

2B: Check the logs — the Activity Log

Open the Activity Log tab.

Every AI action — allowed or blocked — shows up here, newest first.

One row per AI action, with search and filters along the top.

• 1 Search & filters. Find what you need fast — type a name, a tool, or a word from the request; or filter by user, by decision (Allowed / Blocked), or by operation.
• 2 The log. Each row is one action. The columns are explained below.

What each log column shows

Click any row for the full story

Selecting a row opens a detail panel showing everything about that single action:

• Who ran it — name, email, and their Atlassian groups.
• The exact arguments the AI sent (e.g. the search query).
• The decision, and which rule applied if it was blocked.
• The response size and speed (latency).
• An “Atlassian audit JSON” tab showing the same event in Atlassian's native format.

Bonus: The Dashboard — activity at a glance

Open the Dashboard tab for a quick summary over Today, 7 days, or 30 days.

Quick troubleshooting

A few things that occasionally come up — and the one-line fix for each.