Cold-Cache Session Router
A decision layer that chooses how a coding-agent session should resume when its prompt cache has gone cold.
The problem
Resuming a cold-cache coding session can resend a large conversation history and sharply increase token costs. Choosing among different resume strategies depends on the provider, context size, and session state.
Proposed solution
Build a lightweight proxy or CLI that estimates the cost and context trade-offs of available resume strategies, then recommends or applies the best one for that session.
Target audience
Developers and small teams using coding agents across multiple sessions and model providers
Key features
- CLI wrapper for supported coding-agent providers
- Estimate cold-resume token and cost impact
- Compare resume strategies for current context and provider
- Recommend or apply a selected strategy
- Session history and savings report
- Configurable spend and context limits
Monetization strategy
Free local CLI with paid team features or a low-cost hosted dashboard; charge based on seats or managed sessions.
Skills to build it
- CLI and developer-tool design
- LLM provider API integration
- Prompt and context handling
- Usage metering and cost estimation
- Testing across model providers
An idea is a starting point for research. Validate customer demand before building.