Troubleshooting
What each card means when a task is interrupted or a message will not send, and what to do.
Two kinds of card can appear:
- Message not sent: pinned above the composer. The message did not go out and your text is still in the composer; fix the cause and send again.
- Task interrupted: at the end of the reply. Progress is saved, and in most cases "Continue" on the card resumes the run. Once resumed, the card shrinks to a thin divider.
Hover the card title for "Copy error details" and paste the result into a ticket.
Message not sent
| Message | Meaning | What to do |
|---|---|---|
| Not enough team credits | The balance is too low to send | "Manage subscription" to top up or upgrade, see Purchase |
| Your spending limit for this cycle has been reached | The owner's limit for you is used up | Ask the team owner to adjust it, or wait for the next cycle |
| Not available on your current plan | The model, context window or mode is outside your plan | Pick an option within the plan, or upgrade |
| Parallel task limit reached | You have as many running tasks as the plan allows | Wait for a running task to finish |
| A task is running, send again once it finishes | This task is still running | Wait, or type and use "Interrupt the task and send" |
| Host offline, can't dispatch the task | The host is not online, usually your own device is off | Check that the computer is on and online and that "Local connection" in Host Management reads Running |
| Send failed, please try again | A temporary network or service problem | Send again |
Task interrupted
| Title | Meaning | What to do |
|---|---|---|
| Run failed | An error during execution | "Continue" |
| Run interrupted | The execution process exited unexpectedly | "Continue" |
| Failed to start | This turn never started | "Continue" to retry |
| Run timed out | The run reached the 24-hour limit | "Continue" to keep going |
| Run stalled | No output for 10 minutes with no tool running, so the platform interrupted it | "Continue" |
| Cloud computer ran out of resources | A computation exhausted memory or the process limit; the host restarted and your files are safe | "Continue"; ask for large data in chunks and fewer parallel processes or threads |
| Cloud computer hit its process limit | Parallel processes or threads reached the limit; the host restarted and your files are safe | "Continue"; ask for fewer parallel workers, such as n_jobs or thread pool size |
| Context window is full | The conversation reached the model's context limit | Start a new task; workspace files remain available |
| Model service is busy | Upstream rate limiting or overload | Wait a moment, then "Continue" |
| Model service error | An upstream error | "Continue" |
| Model declined this content | Declined by the model's safety policy | Switch to another model and reply "Continue", or start a new task |
| Service updated | The platform released a new version | "Continue" |
| Host offline | Your device disconnected | "Continue" once the host is back online |
| Out of credits | The balance ran out mid-task; the work so far is safe | "Manage subscription" to top up or upgrade, then "Continue" |
| Credit limit reached | Your per-cycle limit is used up | Ask the team owner to adjust it, then "Continue" |
Connection lost
When "Connection lost" appears, the page reconnects as soon as your network is back or you return to it. The task in the cloud was not interrupted, and the timeline fills in.
A permission or question card timed out
Both cards are auto-denied after 10 minutes and ClawsGO continues on that basis. If you want it redone, just send a message saying so.
Still stuck
Open a ticket under Settings → Tickets with the task link and the copied error details; we can see the timeline directly.