I Hit Dia’s Token Limit Mid-Workday: Three Limit Systems Every Atlassian User Should Know

[#1]

I was deep in a Dia work session — pulling context from Slack, Confluence, the usual — when I got this:

"You've hit your token limits. Please try again later."

No warning, no usage meter, just a hard stop. Rovo Chat in Confluence? Worked fine. So this was Dia-specific. I dug in, and what I found surprised me: there are three separate "token limit" systems at play when you use Dia with Atlassian products, and most of us only know about one or two.

The Three Token Limit Systems

1. Dia's Own Free-Tier Usage Limit

This is the one nobody talks about. Dia enforces its own usage caps on Chat, Ask on Page, and Skills — completely separate from Rovo credits. These are enforced by Atlassian on Dia's managed AI pipeline.

• The exact limits are not publicly documented — no per-day cap, no published reset window
• CEO Josh Miller stated the free tier is for users who use AI "a few times a week" (TechCrunch, Aug 2025)
• There is no usage meter and no warning before you hit the wall
• There is no BYOK option — you cannot use your own Claude/Anthropic API key
• Dia Pro ($20/month for unlimited chat) was retired on May 31, 2026. All subscriptions were cancelled and refunded. There is currently no paid plan and no way to upgrade (TidBITS)

2. Rovo Credits (Billing/Consumption)

The one most Atlassian admins already know. Rovo credits are pooled at the org level:

Edition	Per App	Teamwork Collection
Standard	25 credits	250 credits
Premium	70 credits	700 credits
Enterprise	150 credits	1,500 credits

Each chat request = 10 credits. Deep Research = 100. Credits reset monthly, don't roll over. Atlassian is not currently billing for overages and will give 90 days' notice before that changes.

3. LLM Context Window (Technical Limit)

Every AI chat has a maximum token window. Dia pre-loads context from your open tabs, calendar, Slack, and Teamwork Graph connections — so a heavy tab setup can consume most of the window before you type anything.

How to Tell Which One Got You

What You See	Likely Cause	How to Confirm
"You've hit your token limits" + Try Again button	Dia free-tier limit	Rovo Chat in Jira/Confluence still works
Limit hit on a brand new chat	Context window	Close tabs, try again — if it works, context was too large
Credit warning in Atlassian admin	Rovo credits	Check Administration > Insights

What You Can Do Right Now

When Dia's free-tier limit hits:

• Fall back to Rovo Chat in Jira or Confluence — bypasses Dia's limits entirely
• Use the underlying AI model directly — Dia uses Anthropic's Claude models, and a separate Anthropic subscription gives you access without Dia's restrictions
• Wait for the reset — appears to be a rolling daily window based on community reports
• Prioritize Dia for high-value queries — save it for tasks that benefit from connected-tools context

When the context window is the problem:

• Close tab groups — fewer tabs = smaller pre-loaded context
• Start a new chat — each conversation gets its own window

When Rovo credits run low:

• Avoid Deep Research mode (10x the cost of a normal request)
• Wait for the monthly reset

The Real Problem: No Warning

For anyone using Dia as a primary work tool, the "blind death" — a hard cutoff with zero warning — is a real workflow risk. What would help:

• A usage meter or quota indicator
• Warnings before the cutoff — even "you're at 80%" would be enough
• A paid tier or BYOK option to replace the sunset Dia Pro
• Published documentation of the actual limits

How I Researched This

I used Dia itself to investigate — checking Dia's product knowledge base, Atlassian's Rovo documentation, TechCrunch and TidBITS coverage of Dia Pro's launch and sunset, Reddit community threads, and the Dia Pro retirement announcement. I've documented everything in a detailed reference guide on my Confluence space that I keep updated as the situation evolves.

If you've hit this same wall, I'd love to hear your experience in the comments. And if you haven't hit it yet but use Dia for work — bookmark this. With Dia Pro gone and no replacement announced, it's a matter of time.

[Dave Rosenlund _Trundl_]

I can empathize, @John D Patton. I frequently find myself playing AI roulette these days. But it’s extremely frustrating when you’re nearing the finish line and all of a sudden you run out of tokens, and then you have to shift to a different platform.  

Now that we can use Claude agents inside of Atlassian workflows, I’m looking forward to the day when my company upgrades to the enterprise edition of Claude. 

[Dave LIAO]

I hope a Dia product manager sees this and works on increasing transparency on token usage in their tool!