GitHub Repository

Smart(er) code reading for humans and AI agents. Reduces cost per correct answer by ~40% on average. Install: cargo install tilth -or- npx tilth

293 starsRust

Reduce LLM token use by ~30% with this MCP/CLI tool(Claude benchmarked)

Name: Reduce LLM token use by ~30% with this MCP/CLI tool(Claude benchmarked)
Availability: InStock
Author: jahala

by jahala·Mar 5, 2026·2 points·1 comment

Visit Project View on HN

AI Analysis

●●●BangerSolve My ProblemBig BrainSlick

Token-efficient code indexing with adaptive callers tracing cuts Claude costs by 34%.

Strengths

•Measurable ~30-34% cost reduction on real Claude benchmarks, not hand-wavy claims
•Tree-sitter structural search finds definitions reliably, not text grep false positives
•Shared expand budget and session dedup mean multi-query workflows cost less incrementally

Weaknesses

•Narrow audience: primarily valuable if you're already spending heavily on Claude code tasks
•Early-stage: only 53 stars, single maintainer, unproven in production team workflows

Post Description

Smart code reading for humans and AI agents. Tilth is what happens when you give ripgrep, tree-sitter, and cat a shared brain. --

v0.4.4: Added adaptive 2nd-hop impact analysis to callers search — when a function has ≤10 unique callers, tilth automatically traces callers-of-callers in a single scan. First full 26-task Opus baseline (previously 5 hard tasks only). Haiku adoption improved from 42% to 78%, flipping Haiku from a cost regression to -38% $/correct.

v0.4.5: Bumped TOKEN_THRESHOLD from 3500 to 6000 estimated tokens (~24KB), so mid-sized files return full content instead of an outline that agents then read back via 5–7 sequential --section calls. Fixed two major regressions: gin_radix_tree (+35% → ~tie) and rg_search_dispatch (+90% → -26% win). Sonnet hit 100% accuracy (52/52) and -34% $/correct overall.

https://github.com/jahala/tilth/

Full results: https://github.com/jahala/tilth/blob/main/benchmark/README.m...

-- PS: I dont have the budget to run the benchmark a lot (especially with Opus), so if any token whales has capacity to run some benchmarks, please feel free to PR results.