Agent Token Optimizer · Planned OSS release

Audit what your coding agent has to read.

Agent Token Optimizer is being prepared to scan repositories locally for context waste, oversized files, generated content, instruction bloat, editor-indexing risk, token budgets, and model-routing opportunities.

Open example report Release roadmap
Read-only firstThe product direction starts with inspection before repository changes.
Local analysisThe core audit is designed not to require source upload to a hosted service.
Before / after proofMeasure whether a small set of safe changes genuinely improves default context and readiness.

What it is designed to find

Context waste

Large generated artifacts, stale docs, screenshots, logs, and other files that should rarely be default AI context.

Instruction bloat

Oversized agent instructions and opportunities to route tasks into smaller purpose-specific context.

Editor / indexing risk

Generated directories, watcher/search exclusions, large files, and repository shapes that can slow AI-aware editors.

Budget + model routing

Repository-level token budgets and abstract model-tier recommendations for different task types.

See the report direction

Agent Token Optimizer report preview

Public release status

Agent Token Optimizer is planned for public OSS release. The implementation is still being prepared behind a private repository while public packaging, documentation, licensing, and release gates are finalized. Public repository and installation links will appear only when the release is actually accessible.

Product boundary

The project is designed to analyze repository-side context and emit recommendations. It should not treat maximum token reduction as the goal or automatically delete files; recommendations must remain reviewable against the real task and repository.