MCP server and local proxy that compresses LLM prompts, tool output, and replies to cut token cost, with a quality gate that reverts any step that does not save. Exposes llmtrim_compress, llmtrim_compress_text, and llmtrim_stats.
A transparent proxy that sits in front of any other MCP server and shrinks its tool output before it reaches the model. Lossless by default: the transformed bytes are a denser encoding of the same data, with a round-trip gate asserting an exact inverse over the corpus, so nothing is dropped, summarised, or offloaded to a cache that expires. Repeated calls to the same tool emit a delta against the
MCP proxy that compresses tool schemas on the fly. Up to 98% token reduction, 100% signal preserved verified after every compression. Zero LLM calls, fully deterministic.
An adaptive tiny-model layer that sits between an LLM and its MCP tools, compressing verbose tool outputs to reduce token usage by up to two orders of magnitude.
An MCP server that exposes token-optimization pipeline functions as tools, enabling MCP-compatible hosts to reduce token usage in requests before they are forwarded to an Anthropic-compatible backend.
Aggregator MCP proxy that collapses N downstream MCP servers into 4 meta-tools with progressive tool discovery, and compresses large tool outputs (HTML→Markdown, JSON summarization) with full-output retrieval via read_more and a per-session token-savings report.
