token-saver-5000

The semantic compression engine behind gotcontext.ai.

The source-available compression core. It builds an adaptive semantic skeleton of a document or codebase — keeping the load-bearing structure and hiding boilerplate — so you send far fewer tokens without losing the parts a model actually needs.

What it does

• AST-aware code compression that preserves signatures and structure
• Adaptive skeletons: faithful on small docs, aggressive on large ones
• Local ONNX / SBERT / TF-IDF embeddings — no API key, no data leaves the box
• Backs the gotcontext MCP compression tools: ingest_context, read_skeleton, search_semantic, filter_cli_output, gc_pre_flight

Get started

git clone https://github.com/oimiragieo/token-saver-5000
cd token-saver-5000 && uv tool install -e .

token-saver-setup --auto

License: Business Source License 1.1 (source-available)