Observed arrival · 2026-09-22
InferenceBrake Puts a Circuit Breaker on Runaway AI Agents
An open-source system detects reasoning loops, repeated tool calls, and collapsing output patterns before an AI agent keeps spending.
○Open
⊠Login
$Paid
†Ads
✦Pretty
●Pro
◎Niche
◉Human
⚑Risk
ƒJS
- For
- Developers running autonomous AI agents
- Worth noticing
- Six detectors are marked implemented, including compression distance and edit-distance decay.
Field notes
The system watches both generated text and agent actions, allowing it to catch paraphrased repetition, exact token spans, recurring tool calls, and other forms of behavioral stalling. Six methods are labeled implemented, while drift, entropy-collapse, and verifier-model checks remain at different planned stages. The page provides Python, Node.js, and cURL entry points and shows a session-level status check that can break an agent loop.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
○OpenPublic substance visible
⊠LoginAccess appeared gated
$PaidCommerce or pricing visible
✦PrettyNotable craft visible
●ProPolished or operationally mature
◎NicheUnusually specific use
One card from the complete issue