LLM Cost Watchdog
Cost monitoring for AI agents. Tracks pricing across providers, enforces budget ceilings and detects runaway loops before they run overnight.
Intention
An agent loop that keeps calling a model can burn a budget in one night. A watchdog that knows the prices and the loop patterns can stop it.
What exists today
A public MIT-licensed tool and agent skill: price tracking for a large model catalogue, budget ceilings, loop detection through code analysis, and a test suite.
Technical approach
Python. Provider price tables, budget rules, and AST analysis of agent code to flag loops that call models without a stop condition.
Evidence and known limitations
- Price tables go stale between updates.
- Loop detection is heuristic and can miss or over-flag.