Grok Bot Cloud Agent consistently fails with ResourceExhausted — vttablet transaction pool connection limit exceeded

[Sanghoon_Kang]

Where does the bug appear (feature/product)?

Grok Bot

Describe the Bug

Hi Cursor team,

I’m opening this as a separate bug report following Colin’s suggestion in my previous Grok Bot thread.

Since yesterday, my Grok Bot has been unable to launch a delegated Cloud Agent because of the following error:

ResourceExhausted — vttablet transaction pool connection limit exceeded

Latest Caller ID:
jqftxh24a9mn90ceno1y

This does not appear to be my plan usage limit.

I upgraded to Cursor Ultra ($200/month), and my current usage is still very low:

Cursor Models: ~6% used
Other Models: ~1% used
Grok Bot weekly usage: ~5%
On-demand spending: disabled
The workflow itself is ready and authorized, but the Cloud Agent cannot launch. We have retried at widely separated times, including again today, and the same ResourceExhausted / vttablet transaction pool error persists.

The parent/Builder correctly stops retrying and reports the task as PLATFORM BLOCKED, so there have been no repository changes or repeated retry loops.

Could you please confirm whether this is:

a Cursor-side Cloud Agent infrastructure/capacity issue,
an account/workspace concurrency or connection-pool limit,
or another Cloud Agent quota/limit that is separate from the Ultra usage shown in Plan & Usage?
If there is a way to inspect or reset the relevant Cloud Agent transaction pool for my account/workspace, that would also be very helpful.

The important distinction for me is whether I should simply wait for platform capacity to recover, or whether there is an account/workspace condition that needs intervention.

Thanks!

Steps to Reproduce

1. Open Grok Bot and use a parent/orchestrator Bot that delegates an implementation task to a project Builder Bot.

2. Have the Builder launch a delegated Cursor Cloud Agent for an authorized implementation task.

3. The Cloud Agent launch fails with:

ResourceExhausted — vttablet transaction pool connection limit exceeded

4. Wait and retry later using the same authorized task. The same ResourceExhausted error occurs again.

5. In my case this has persisted across widely separated attempts since the previous day, not just rapid consecutive retries.

6. Verify Plan & Usage. My account is on Cursor Ultra ($200/month) and usage is far below the included limits:

• Cursor Models: approximately 6% used
• Other Models: approximately 1% used
• Grok Bot weekly usage: approximately 5%
• On-demand spending: disabled

7. The Builder therefore cannot launch the Cloud Agent and correctly stops without modifying the repository.

Latest error:
ResourceExhausted — vttablet transaction pool connection limit exceeded

Latest Caller ID:
jqftxh24a9mn90ceno1y

Expected Behavior

A delegated Cloud Agent should launch normally while the account has available included Ultra usage and there is no product/workflow authorization blocker.

If the Cloud Agent cannot launch because of a separate infrastructure, concurrency, transaction-pool, or account/workspace limit, the product should clearly identify that condition and distinguish it from normal plan usage exhaustion.

If this is a temporary Cursor-side capacity issue, the launch should succeed once capacity is available again.

If there is an account/workspace-specific Cloud Agent connection or transaction-pool limit, the UI should expose the relevant limit/status or provide a clear recovery path.

Screenshots / Screen Recordings

Operating System

MacOS

Version Information

버전: 0.53.0
릴리스 트랙: stable
OS: darwin

For AI issues: which model did you use?

Auto

The issue occurs while Grok Bot / Builder is attempting to launch a delegated Cursor Cloud Agent. The failure occurs at Cloud Agent launch with ResourceExhausted, rather than during model inference inside the task.

For AI issues: add Request ID with privacy disabled

No bc-… Background Agent ID was created because the Cloud Agent failed during launch.

Latest Caller ID:
jqftxh24a9mn90ceno1y

Error:
ResourceExhausted — vttablet transaction pool connection limit exceeded

Additional Information

I am filing this as a separate bug report following a suggestion from Cursor Community Support Engineer Colin in my previous Grok Bot thread.

This does not appear to be normal plan usage exhaustion.

Current plan:
Cursor Ultra — $200/month

Current usage at the time of the issue:

• Cursor Models: ~6%
• Other Models: ~1%
• Grok Bot weekly usage: ~5%
• On-demand spending: disabled

The same Cloud Agent launch failure has persisted across attempts separated by many hours.

The workflow itself is not blocked by Product approval or repository state. The task is authorized and ready to execute.

The Builder reports:

• no SHA/build produced
• repository unchanged
• no automatic retry loop
• execution blocked specifically at Cloud Agent launch

I would particularly like to know whether the vttablet transaction pool limit is:

1. Cursor-side infrastructure/capacity,
2. an account/workspace concurrency or connection-pool limit,
3. or a separate Cloud Agent quota not represented in Plan & Usage.

If this is account/workspace-specific, please let me know whether the relevant transaction pool can be inspected or reset.

Does this stop you from using Cursor

Sometimes - I can sometimes use Cursor

[Sanghoon_Kang]

Update: the issue has now recovered without any account or plan changes.

After the ResourceExhausted / vttablet transaction pool error persisted across multiple attempts for roughly a day, I retried the exact same authorized Grok Bot → Builder → Cloud Agent workflow today.

This time the Cloud Agent launched successfully and the workflow completed end-to-end:

• Cloud Agent launch: SUCCESS
• Implementation completed
• Commit/push completed
• Independent QA completed
• 49/49 independent attacks passed
• E2E 219/219 passed

I did not enable on-demand spending or make any additional plan/usage changes before the successful retry.

So the issue appears to have been transient or infrastructure/capacity-related rather than caused by exhaustion of my Ultra plan usage.

I’m leaving this report open because the ResourceExhausted condition persisted for many hours and there was no indication in the UI of whether/when the Cloud Agent transaction pool would recover.

It would still be helpful to know what caused the vttablet transaction pool limit and whether users can distinguish this condition from account usage/quota exhaustion.

Thanks.

[kevinn (Kevin Neilson)]

Hey @Sanghoon_Kang, thanks for the report, and for the update that the same workflow completed.

To answer your three questions: this was a Cursor-side service degradation, not anything on your account. On Sep 17 between 15:20 and 20:04 UTC (Sep 18, 00:20 to 05:04 KST) Cloud Agents and Automations were degraded: Cursor Status - Investigating service degradation — Cloud Agents and Automations

Cloud Agent launches during that window were failing with that same “transaction pool connection limit exceeded” message. It is not a per-account pool, not a concurrency limit, and not a separate quota, and there is nothing on your side to inspect or reset. Your Ultra usage was never the blocker. The “Caller ID” in the message is an internal identifier, not something tied to your workspace.

I checked your account: the launch attempts on Sep 18 between 00:46 and 01:44 KST hit that window. The next launch that reached us was Sep 19 at 00:58 KST, and that one succeeded. That matches what you reported: the workflow completed without any plan or account change. Earlier “still failing” updates look like the Builder repeating its PLATFORM BLOCKED card, which says it will not launch again until a human re-hands the task.

If a launch fails again, reply here with the time it happened (with timezone) and the exact message, and I’ll take another look.