Forgive the double post - also in "help
Forgive the double post - also in "help" for Krill:
I'm tying myself in knots trying to work out why I'm furiously burning tokens, even when not using clawdbot. Have exhausted my daily Claude max, Codex, and some Openrouiter credits on top. Heartbeat set to every two hours, but Clawdbot digging around seems to think it is in some sort of carzy burn loop and is being called in the system prompt. Any suggestions welcome. Clawdbot on mac mini, node on MacOS, mainly using Telegram to communicate to clawdbot. in case helpful, here's one of the reported investigations:
ROOT CAUSE IDENTIFIED
The "heartbeat prompt" is being automatically appended to system event messages (like "Exec completed", "Node connected", etc.)!
When the system sends notifications like:
• System: [timestamp] Exec completed (lucky-wi, code 0) ...
• System: [timestamp] Node: Tim's MacBook Pro ...
Those system messages are followed by the heartbeat prompt text. And each one triggers a full heartbeat run.
This means:
• Every time a background exec completes → heartbeat prompt → agent run
• Every time a node connects → heartbeat prompt → agent run
• Every time there's a system event → heartbeat prompt → agent run
That's why there were so many heartbeat runs!
Where is this coming from?
This is either:
1. Clawdbot's internal system - the gateway/session manager is automatically appending the heartbeat prompt to system events
2. A channel plugin/automation that's configured to send heartbeat checks after system events
-----
Anyone got any tips?
I'm tying myself in knots trying to work out why I'm furiously burning tokens, even when not using clawdbot. Have exhausted my daily Claude max, Codex, and some Openrouiter credits on top. Heartbeat set to every two hours, but Clawdbot digging around seems to think it is in some sort of carzy burn loop and is being called in the system prompt. Any suggestions welcome. Clawdbot on mac mini, node on MacOS, mainly using Telegram to communicate to clawdbot. in case helpful, here's one of the reported investigations:
ROOT CAUSE IDENTIFIED
The "heartbeat prompt" is being automatically appended to system event messages (like "Exec completed", "Node connected", etc.)!
When the system sends notifications like:
• System: [timestamp] Exec completed (lucky-wi, code 0) ...
• System: [timestamp] Node: Tim's MacBook Pro ...
Those system messages are followed by the heartbeat prompt text. And each one triggers a full heartbeat run.
This means:
• Every time a background exec completes → heartbeat prompt → agent run
• Every time a node connects → heartbeat prompt → agent run
• Every time there's a system event → heartbeat prompt → agent run
That's why there were so many heartbeat runs!
Where is this coming from?
This is either:
1. Clawdbot's internal system - the gateway/session manager is automatically appending the heartbeat prompt to system events
2. A channel plugin/automation that's configured to send heartbeat checks after system events
-----
Anyone got any tips?
The AI that actually does things. Emails, calendar, home automation - all from your favorite chat app. New shell, ready to help. The lobster way.
131,281Members
Resources
Similar Threads
Was this page helpful?
