Tanuki Context claims up to 94% token savings by turning some LLM inputs into images
0syna · reddit · 2026-07-27
Tanuki Context is an open-source MCP server and proxy that tries to cut LLM context costs by turning some inputs into images when that is cheaper than sending text.
The author claims up to 94% token savings on logs and describes several techniques:
- a log distiller that collapses repeated lines while keeping errors verbatim
- a columnar codec for JSON so keys are stated once
- a cost model that prefers cache reads over re-imaging when cached tokens are cheaper
It works with vision-capable models and supports Claude Code, the Claude Agent SDK, and a machine-wide proxy mode via ANTHROPICBASEURL. The author says output tokens are unaffected and retrieval is still cheaper for some narrow queries.
More from coding & agent
- OpenAI splits Codex and ChatGPT Work into code-first and knowledge-work products — johnseach · 2026-07-27
- Memoria open-sources Git-style version control for AI agent memory — bigaiguy · 2026-07-27
- Open-source jcode claims 14 ms startup and multi-agent repo coordination — thisguyknowsai · 2026-07-27
- New ComfyUI prompt manager node adds on-the-fly edits, randomization, and shared presets — 3deal · 2026-07-27
- VS Code Remote makes the Claude-and-Codex server setup usable anywhere — nico_jeannen · 2026-07-27
- Developer runs Claude and Codex on a server to work from a phone or iPad — nico_jeannen · 2026-07-27