Compress prompts 40-60% using local LLM + embedding validation. Preserves all conditionals.
io.github.base76-research-lab/token-compressor
https://github.com/base76-research-lab/token-compressor
STDIO
No auth required
Hosted endpoint — paste into any MCP client.
Where to find authoritative docs and source for token-compressor.
Open MCP Agent Studio and connect this server to Claude, GPT, Gemini, DeepSeek and more — no install required.
Open Agent Studio