ChatGPT Connectors for Jira and Other Trackers

The ChatGPT connector for Jira is Atlassian’s own app, called Atlassian Rovo. You add it from ChatGPT’s Plugin Directory, sign in with your Atlassian account and approve, with no developer mode needed. ChatGPT can then search Jira, summarise work items and create new ones, and it does all of it with your own Jira permissions. The same app carries Confluence. Several other trackers have official listings too: Trello, Asana, monday.com, Notion and Linear among them.

A word on names, because they have changed more than once. OpenAI started with “connectors”, moved to “apps”, and its newer menus and developer documentation say “plugins”. They are the same thing: a listed integration you connect once and then call from a chat. This post uses “app”, and says “connector” where a vendor does.

The Jira app: Atlassian Rovo

Atlassian Rovo in ChatGPT is built on Atlassian’s remote MCP server, the same one Claude, Cursor and VS Code use. Atlassian’s MCP documentation↗ (opens in a new tab) lists OpenAI ChatGPT among its supported clients and describes the server as working only with Atlassian Cloud apps, so Jira Data Center and Server are not covered.

When Atlassian announced the ChatGPT connector↗ (opens in a new tab) in December 2025, it named three jobs: summarise Jira work items and Confluence pages inside ChatGPT, create Jira issues from the conversation, and automate multi-step work such as generating or updating issues in bulk. Underneath are the Jira tools of the MCP server, grouped as read, write and search, with delete and project management switched off until an Atlassian admin enables them. The full tool list is in what the Jira MCP server can do.

Connecting it

1. In ChatGPT, open the Plugins tab to browse the Plugin Directory↗ (opens in a new tab). Older versions kept the same list under Settings, as Apps & Connectors.
2. Search for Atlassian Rovo, open its details and select the plus button to install it, then connect it when prompted.
3. Your browser opens on Atlassian’s sign-in. Sign in with the account you use for Jira.
4. Read the consent screen: which app is asking, which site, and what it may do. Approve if it matches what you expect.
5. Back in ChatGPT, start a new chat and name Jira in your message, or type @ and pick Atlassian Rovo, so ChatGPT uses the app instead of guessing.

If the app is missing or greyed out, the cause is usually on one of two sides. On OpenAI’s side, what you can connect depends on your plan, your region and, in a company workspace, your admin. On Atlassian’s side, admins decide which AI clients may sign in to their organisation, as the next section explains.

Using it in a chat

Start with reads, so you can see how ChatGPT understands your projects before it changes anything:

• “Which Jira projects can you see, and who am I signed in as?”
• “Find the open bugs in PAY assigned to me, newest first, and show me the JQL you used.”
• “Summarise PAY-412 with its comments, and tell me what is blocking it.”

Then one write you can check by eye in Jira: “Create a bug in PAY titled ‘Retry job double-charges on timeout’ with these steps to reproduce.” For bulk work, such as turning a meeting’s notes into a dozen work items, ask for the list first, correct it, and only then say “create them”. Ask it to name every key it created, so you can open each one.

Who controls what

On the Jira side, the app can do nothing you could not do yourself: Atlassian says every action respects the signed-in user’s existing permissions. Atlassian admins decide which AI clients may use OAuth against their organisation, by domain; Atlassian’s list of supported domains↗ (opens in a new tab) includes chatgpt.com, and an admin can remove it. Admins can also filter the organisation’s audit log for Rovo MCP actions.

On the ChatGPT side, OpenAI’s Help Center describes per-app permission settings with options such as always ask, allow read actions, allow low-risk actions and allow all actions. By default ChatGPT reads without asking and stops before actions that could have a meaningful effect outside ChatGPT, expose sensitive information or be hard to undo. These settings never widen what Jira allows; they only change when ChatGPT checks with you. For Jira, allowing reads and keeping writes on ask is a sensible start.

In a ChatGPT Business, Enterprise or Edu workspace, admins also decide which apps members may use at all. If Atlassian Rovo shows as disabled, that is the person to ask.

Other trackers with official ChatGPT listings

These are the ones whose makers document a ChatGPT app or connector on their own sites. Each works through your own account on that product, so the product’s permissions are the outer limit.

• Trello. Atlassian’s other tracker has its own listing, separate from Atlassian Rovo, and connects one Trello workspace at a time. The steps and limits are in Trello MCP with ChatGPT.
• Asana. Asana’s ChatGPT page↗ (opens in a new tab) says you connect it from ChatGPT’s app directory, now the Plugin Directory, and then mention Asana in a chat to create projects, assign tasks and check status. An admin may need to allow it first.
• monday.com. The monday.com developer documentation↗ (opens in a new tab) lists a monday connector in OpenAI’s ChatGPT among the clients that work with its MCP server.
• Notion. Notion has a ChatGPT app, and its MCP server can also be added through developer mode; Notion MCP with ChatGPT compares the two routes.
• Linear. Linear’s ChatGPT integration page↗ (opens in a new tab) still describes a research connector, but ChatGPT now lists Linear as a plugin that finds and references issues and projects, keeping each user’s permissions. Linear in ChatGPT covers what it can change.

The list changes month by month, and availability depends on your ChatGPT plan, region and workspace. Search the Plugin Directory in your own account before you assume a tracker is missing.

When there is no listing: developer mode

If a tracker publishes a remote MCP server but has no ChatGPT listing, developer mode lets you add the server’s address as your own app, with OAuth sign-in, on the plans OpenAI lists for it. OpenAI describes developer mode as powerful but risky, meant for people who understand prompt injection and mistaken writes, and on company plans it may be limited to admins. The step-by-step version, using Trello as the example, is in Trello MCP with ChatGPT; the reasons to be careful are in MCP security risks.

What these apps have in common

• They act as you. Jira, Asana and the others see your account, with your full rights inside whatever you approved.
• There is no separate identity for ChatGPT. A work item it created looks like one you created, so agree a habit, such as a line in the description, if your team needs to tell them apart.
• The tracker sets the ceiling and ChatGPT sets the prompts. Neither side’s settings grant anything the other refuses.
• Admins on both sides can switch them off, which is worth knowing before you build a routine around one.

A board that ChatGPT joins as a member

fenbs takes the opposite approach to identity. It is a small board with four lanes, To Do, Next Up, In Progress and Completed, and three kinds of task, feature, enhancement and bug. ChatGPT connects to it as a developer-mode app, ChatGPT’s name for a custom MCP server, where your plan offers developer mode, with a browser sign-in and nothing to copy. It works with your role on the board, narrowed by the scopes you tick when you approve it, or with a narrower role of its own, and every change it makes is recorded in the history with its name. You can revoke it under Settings without touching your own sign-in. The steps are on connecting ChatGPT to fenbs.