Are there limits set for Agent Mode in Databricks Free Edition?

[wirrywoo]

Hi everyone,

I’ve been trying out Agent Mode in the Databricks Free Edition and I’m a bit unclear on the usage limits.

From what I can tell, Agent Mode seems to default back to chat mode after a certain point, and when I was using it more heavily, I started noticing what looks like resource throttling (responses slowing down or not completing as expected). That makes me think I may have hit some kind of limit.

I understand Agent Mode is more resource-intensive since it runs multi-step tasks, but I haven’t been able to find clear documentation on what limits are actually enforced in the free tier (e.g. number of runs, compute usage, or query caps).

I’d really like to continue using Agent Mode since it’s been genuinely helpful for my workflow so far, so I’m hoping this is just a temporary limit rather than a hard stop.

Does anyone know:

• Are there specific limits for Agent Mode in the free edition?
• Do these limits reset (daily or monthly), or are they fixed for the duration of the trial?
• Is the fallback to chat mode expected behavior after hitting a limit?

Would really appreciate any insight from people who’ve run into this or have more details.

Thanks!

[balajij8]

Databricks workspace's compute resources will be shut down and unavailable for the rest of the day (a month in extreme cases) if you exceed the quota. While the compute will be unavailable, Data and settings will not be deleted. You can resume the work when the limit resets. Free Edition accounts may not be used for commercial purposes

More details here

[wirrywoo]

Thanks for the response. I don’t think this is due to compute becoming unavailable, since I’m still able to run code without any issues. I’m using the Databricks Free Edition strictly for personal side projects, and I’ve been working with Genie in Agent mode to help guide me in writing future-proof code.

[Ashwin_DSA · Accepted solution]

Hi @wirrywoo,

The page shared by @balajij8 is the best resource to understand the limitations in Free Edition. Resets daily under normal conditions. 

In Free Edition there isn’t a separate published "Agent Mode quota", but Agent Mode does run on a more constrained free-tier LLM/Genie pool than the notebooks themselves. That pool has soft per-workspace limits (for example, Genie supports about 20 questions per minute via the UI and around 5 per minute via the API in the free tier), and agentic flows tend to use that capacity faster because they do multi-step reasoning and retries behind the scenes. You may also want to check this page for Genie limits which is applicable to both Free and Paid workspaces.

When those LLM guardrails kick in, Databricks intentionally degrades Agent Mode rather than shutting down your workspace. You’ll see slower or partial answers and, in some cases, the UX will fall back to chat mode even though your regular compute (running code in notebooks, etc.) still works. That’s why it feels like only Agent Mode is affected.

These are soft, rolling limits, not a permanent flag. The per-minute limits constantly reset, and the overall Free Edition fair-usage quota resets over time as well. There is also some ongoing backend work right now to tighten abuse protections for agentic features in Free Edition, which may temporarily make Agent Mode less available there than in full Databricks trials.

If you want sustained, heavy use of Agent Mode (for example, letting it drive a lot of multi-step coding tasks), the most reliable path today is to do that in a regular Databricks trial workspace, and reserve Free Edition for lighter, occasional agentic use and normal notebooks/SQL.

If this answer resolves your question, could you mark it as “Accept as Solution”? That helps other users quickly find the correct fix.