Vexp – Local context engine reducing AI coding costs by 58%

Summary: Vexp pre-indexes codebases into a dependency graph and delivers only relevant context to AI coding agents, cutting token usage and tool calls significantly. It runs entirely locally without cloud or accounts, improving efficiency for tools like Claude Code and Cursor.

What it does

Vexp builds a dependency graph of your codebase using Rust, tree-sitter, and SQLite, then serves ranked context via a single MCP call to AI agents. This prevents unnecessary file exploration and reduces token waste during code generation.

Who it's for

It is designed for developers using AI coding agents who want to optimize token consumption and reduce redundant tool calls during coding tasks.

Why it matters

By focusing AI agents on relevant code context, Vexp lowers costs by 58%, decreases output tokens by 63%, and minimizes tool calls by 90%, enhancing coding efficiency without cloud dependencies.