Critical Token Leak & Context Bloating in Moltbot (January 2026 Implosion)
I am experiencing a massive, uncontrolled spike in token usage with my Moltbot (formerly Clawdbot) instance over the last 48 hours. The bot has become unresponsive to instructions and appears to be "looping" or sending massive amounts of redundant data.
Key Statistics (from my Anthropic Usage Dashboard):
• Total Tokens In: 481,624,431
• Total Tokens Out: 1,750,222
• Input-to-Output Ratio: ~275:1
• Timeframe: 2 days
Issues Observed:
• Extreme Context Bloating: Every small prompt seems to trigger a re-send of the entire conversation history and all attached files, leading to astronomical input costs.
• Instruction Ignored: The bot is ignoring system prompts and CLAUDE.md configurations, behaving erratically and failing to complete simple tasks.
• Potential Security Breach: I suspect this might be related to the reported "January 2026 Moltbot Implosion," where public instances and API keys are being targeted or leaked.
Questions for the community:
1. Is this a confirmed bug in the Moltbot agent logic or the Anthropic API handling for January 2026?
2. How can I force the bot to use a "Sliding Window" or "Prompt Caching" to stop it from re-sending the entire context every time?
3. Are there any specific security patches I should apply to prevent unauthorized token consumption?
I have already rotated my API keys, but the consumption resumes as soon as the bot starts a new session. Any help would be greatly appreciated before my budget is completely drained.
Key Statistics (from my Anthropic Usage Dashboard):
• Total Tokens In: 481,624,431
• Total Tokens Out: 1,750,222
• Input-to-Output Ratio: ~275:1
• Timeframe: 2 days
Issues Observed:
• Extreme Context Bloating: Every small prompt seems to trigger a re-send of the entire conversation history and all attached files, leading to astronomical input costs.
• Instruction Ignored: The bot is ignoring system prompts and CLAUDE.md configurations, behaving erratically and failing to complete simple tasks.
• Potential Security Breach: I suspect this might be related to the reported "January 2026 Moltbot Implosion," where public instances and API keys are being targeted or leaked.
Questions for the community:
1. Is this a confirmed bug in the Moltbot agent logic or the Anthropic API handling for January 2026?
2. How can I force the bot to use a "Sliding Window" or "Prompt Caching" to stop it from re-sending the entire context every time?
3. Are there any specific security patches I should apply to prevent unauthorized token consumption?
I have already rotated my API keys, but the consumption resumes as soon as the bot starts a new session. Any help would be greatly appreciated before my budget is completely drained.
The AI that actually does things. Emails, calendar, home automation - all from your favorite chat app. New shell, ready to help. The lobster way.
134,827Members
Resources
Similar Threads
Was this page helpful?
