ContextOps: a free offline tool that scores prompt token waste
ContextOps is an open-source tool that checks a prompt's structure before it's sent to a model and gives it a Context Health Score from 0 to 100. In and AI agents, prompts quietly accumulate waste over time: duplicated chunks of retrieved text, bloated , oversized , and repeated tool outputs.
This waste drives up cost and can make model behavior less consistent, but it's hard to spot until it becomes a real problem. ContextOps flags issues like redundancy, token waste, structural imbalance, and over-reliance on a single source.
The scope is deliberately narrow: it makes no calls to AI models, uses no embeddings, needs no API keys, runs entirely offline, and always produces the same score for the same input. It intentionally does not try to judge prompt quality, reasoning ability, semantic similarity, or (made-up facts).
Key points
- Analyzes prompt structure before it reaches the model and outputs a 0-100 Context Health Score
- Flags duplicated , bloated , oversized , and repeated tool outputs
- Runs fully offline with no model calls, no embeddings, and no API keys, giving results
- Deliberately does not evaluate prompt quality, reasoning, semantic similarity, or