Tobler

Overview

Tobler is a Python library and MCP server that claims up to 99% reduction in read token usage for AI coding agents by using query-adaptive context assembly instead of fixed-size chunk retrieval.[3] Inspired by Tobler’s first law of geography, it dynamically decides how much code to feed the agent based on the query, making LLM-based coding more efficient and cheaper.[3] Developers should care because it slots into existing MCP-compatible tools and materially lowers context costs for large codebases.