tokenfold

View on GitHub

Private, provider-neutral context compression for LLMs. Reversibly compress schemas and logs inside your perimeter with zero third-party calls.

★ 13135 forksRustApache-2.0snchimata

Tokenfold compresses structured JSON, tool schemas, and logs before they reach an LLM. Its Rust core performs deterministic, reversible folding; optional recoverable pruning and a local proxy, CLI, Python, TypeScript, and MCP interfaces support integration.

Use Cases

Reduce token usage for structured JSON sent to LLMsCompress tool schemas in agent workflowsOptimize structured RAG feedsRoute OpenAI-compatible requests through a local compression proxy

Built With

Language: Rust
Frameworks: LangChain · LlamaIndex · LiteLLM · Instructor

tokenfold

Use Cases

Built With

Tags