How to use Claude to reduce code lookup tokens

Indexio began with its authors watching Claude Code's built-in tool calls. They noticed substantial time going into regex searches for lines of code and investigated codebase indexing as an alternative. Their concern was that repeated regex results brought unnecessary data into the context window while the underlying files were mostly static.

They built Indexio to index a codebase into an internal database and return requested information through MCP. Getting sessions to use the indexed lookup required additional tool-call logic that attempts to override built-in lookup tools. When information is absent from the index, the system falls back to regex searches. The implementation is a single Rust binary.

The authors reported roughly 50–60% lower token usage in their real-world testing compared with built-in tools, alongside near-instant lookups rather than multiple seconds for each regex lookup. Those are the authors' results; the post supplies no detailed measurement procedure. They also identified a weakness: symbol lookups can work less well than regex, limiting the fit for symbol-heavy workflows. Who it's for: developers evaluating indexed code discovery for agents working repeatedly in an established codebase.

Prompt
Suggested prompt (written by Agents In Use, not the author's)

Source: Hacker News · iohotspot