Jira Automation: “Use Agent” always returns Bad Request (400), while “Use Rovo” works

[mhaas]

Hi,

I’m experiencing an issue with the Use Agent action in Jira Automation.

Every execution of Use Agent fails with:

Error returned from Rovo
Bad Request (400)

This happens regardless of:

• which Rovo agent I select;
• which prompt I use;
• whether I use smart values;
• whether I include the Jira issue description;
• whether the agent has skills or knowledge sources.

The following test works successfully:

• The same automation flow;
• The same prompt;
• The same rule actor;
• Use Rovo instead of Use Agent.

The audit log shows:

Use Rovo
Successfully generated response for prompt

Use Agent
Error returned from Rovo
Bad Request (400)

I have already tried the following:

• creating a completely new automation flow;
• removing and recreating the Use Agent action;
• testing multiple agents;
• testing a simple prompt such as Reply with OK only;
• removing all smart values;
• using an agent without skills or knowledge sources;
• using my own account as the rule actor;
• enabling the agent’s Jira work-item collaboration option.

The issue started recently. The same functionality worked last week, and no changes were made to the automation rule or agent configuration.

Question:
Is the Use Agent action currently affected by a known Jira Automation/Rovo issue? Has anyone else experienced HTTP 400 errors for every agent while the Use Rovo action continues to work?

I would especially appreciate advice on:

1. where to find more detailed error information or a correlation ID;
2. whether a specific permission or entitlement is required for Use Agent;
3. whether Use Agent has recently been changed, deprecated, or migrated;
4. whether this should be reported as an Atlassian Cloud bug.

Thanks!

Additional troubleshooting / minimal reproduction

I have now reproduced this with a completely new Rovo agent.

Agent instruction: This is a test agent. Reply with: Hallo Mariska

The agent works correctly when invoked directly through Rovo Chat.

The same agent immediately fails when invoked through Jira Automation → Use agent:

Error returned from Rovo
Bad Request (400)

No knowledge sources, skills, Confluence searches or complex prompts are involved in this reproduction.

In the same automation execution, the Use Rovo action succeeds, while Use agent returns HTTP 400.

Site subscriptions:

• Jira: Standard
• Confluence: Premium
• Rovo: included/free app with included Rovo Credits

I am also an Organization Admin.

This appears to isolate the issue to the Jira Automation → Rovo Agent invocation rather than the agent configuration itself.

Could someone from Atlassian please check whether this is a known regression or escalate this internally? In particular, it would be useful to know why the Rovo invocation endpoint is returning HTTP 400 for Use agent.

[Aditya Srivastav]

Hello,

Based on the troubleshooting you’ve already done, this does look like an issue with the Jira Automation → Rovo Agent integration, rather than the agent configuration itself.

The strongest indicators are:

• Use Rovo works in the same automation with the same prompt and rule actor.

• A new Rovo agent with only a simple instruction also fails.

• The agent works correctly when invoked directly in Rovo Chat.

• The failure is consistently HTTP 400 (Bad Request) specifically from Use agent.

• Removing smart values, skills, knowledge sources, and complex prompts does not change the result.

That effectively isolates the problem to the Automation-to-Agent invocation path.

I would recommend reporting this to Atlassian Support as a potential Cloud regression/bug and include the exact timestamp of a failed execution, rule ID, agent ID/name, site URL, and screenshots of the Automation audit log.

Also ask Support to check the backend logs for the Rovo request/correlation ID associated with the 400 response. The Automation audit log does not always expose enough diagnostic information for this type of failure.

I would not spend more time recreating agents or changing prompts at this point. Your minimal reproduction is already strong evidence that the agent itself is functioning correctly.

For the support ticket, I would summarize it as:

“Rovo Agent works normally in Rovo Chat but consistently returns HTTP 400 when invoked through Jira Automation → Use Rovo agent. Use Rovo succeeds in the same automation. The issue reproduces with a newly created agent containing only a static instruction and no skills, knowledge sources, or smart values.”

This should give Atlassian enough information to investigate the backend integration directly.

[mhaas]

Thanks Aditya, this confirms my conclusion as well.

Unfortunately, I’m currently unable to raise a ticket with Atlassian Support. I’m an Organization Admin, and our site has Jira Standard and Confluence Premium, but when I try to contact Atlassian Support I’m redirected to the Community.

Because of that, I’m currently unable to provide this minimal reproduction directly to Atlassian Support.

Do you know if there is a way to have this escalated to Atlassian from the Community, or could an Atlassian team member take a look at this thread?

The issue is still consistently reproducible, including with a brand-new agent whose only instruction is to return a static response.

[mhaas]

Update: I was able to raise a support ticket after all. The support portal's AI chat initially redirected me to the Community, but after clearly stating this was a reproducible bug (not a how-to question) and asking to escalate, it let me submit a request to a human support agent.

For anyone hitting the same "redirected to Community" wall on a Standard plan: go to the support portal, choose "Technical issues and bugs", select Rovo as the app, and explicitly state it's a bug that reproduces with a minimal example. Ask to escalate to a human and submit a support request.

The support agent confirmed this points to a likely regression in the automation invocation path and asked me to submit the request with the rule ID and a failed-run timestamp so engineering can locate the 400 in the backend logs. Ticket is now filed.

[Aditya Srivastav]

Glad you were able to get the ticket through. The fact that Support also sees this as a likely regression in the Automation → Rovo Agent invocation path and has requested the rule ID and failed-run timestamp is a strong indication that the issue is on the service side rather than with your agent configuration.

Hopefully the backend logs will give them enough information to identify the 400 response and get this resolved.