← All comparisons
LemonCrow vs CodeGraph

What CodeGraph says, vs. what it scored.

Local SQLite knowledge graph of symbols, call edges, and dependencies, built via tree-sitter and queried over MCP.

What CodeGraph says about itself
“Surgical context, fewer tool calls, faster answers, 100% local.”
“CodeGraph hands the agent the exact code it needs in one call. It's a pre-built knowledge graph of every symbol, call edge, and dependency in your codebase.”
Publishes some numbers ...never against another search tool View source ↗
What it actually scored — same 14 repos, same 7,213 queries as every other tool
Tool MRR p95 p100
LemonCrow +semantic (BGE) 0.727 390ms 1057ms
LemonCrow lexical (default) 0.676 134ms 319ms
CodeGraph 0.296 17ms 532ms

Own numbers real, fair with/without comparison. Fastest raw latency measured -- 17ms p95. 0.296 MRR vs. LemonCrow's 0.727.

By query kind -- same benchmark, broken out (no reps in this eval: one deterministic pass per query)
Kind LemonCrow +semantic LemonCrow lexical CodeGraph
definition 0.873 (n=1570) 0.871 (n=1570) 0.766 (n=1570)
content 0.873 (n=1444) 0.864 (n=1444) 0.000 (n=1444)
semantic 0.759 (n=1800) 0.576 (n=1800) 0.051 (n=1800)
swebench 0.500 (n=1908) 0.493 (n=1908) 0.364 (n=1908)
sessions 0.587 (n=491) 0.571 (n=491) 0.299 (n=491)

n = query/gold pairs of that kind, out of 7,213 total -- every provider scored on all 5 kinds.

By repo -- all 15 repos in the corpus, same query set
Repo LemonCrow +semantic LemonCrow lexical CodeGraph
astropy/astropy 0.772 0.715 0.310
django/django 0.689 0.652 0.286
lemoncrow-lab/lemoncrow-dev 0.467 0.477 0.304
lemoncrow/lemoncrow 0.594 0.557 0.294
matplotlib/matplotlib 0.801 0.747 0.279
mwaskom/seaborn 0.814 0.768 0.331
pallets/flask 0.735 0.671 0.304
psf/requests 0.840 0.803 0.373
pydata/xarray 0.815 0.764 0.354
pylint-dev/pylint 0.856 0.784 0.359
pytest-dev/pytest 0.826 0.739 0.322
scikit-learn/scikit-learn 0.740 0.669 0.264
sphinx-doc/sphinx 0.637 0.580 0.225
sympy/sympy 0.694 0.637 0.283
torvalds/linux 0.726 0.668 0.158

MRR per repo: n-weighted blend across all 5 query kinds, same 7,213-query run.

Cost, head-to-head -- a different suite: 7 repos, 5 reps each, claude-opus-4-8
Repo CodeGraph vs baseline
Tokio $3.44 28.1% pricier
Alamofire $2.48 48.6% cheaper
Django $2.32 even
OkHttp $1.35 15.6% cheaper
VS Code $2.56 16.6% cheaper
Gin $1.36 24.9% pricier
Excalidraw $2.47 29.6% cheaper
Total $15.99 16.3% cheaper

Measured on LemonCrow's own 7-repo, 5-rep exploration-task suite (claude-opus-4-8), not CodeGraph's self-published with/without numbers. Cheaper than baseline on 5 of 7 repos (16-49%), pricier on Tokio (+28%) and Gin (+25%), and 22.7% slower overall despite the lower total cost -- a noisier result than CodeGraph's own published claims, which show it uniformly faster and never pricier on this same repo set.

Baseline total: $19.11. Wall-clock: 22.7% slower than baseline overall. Full per-repo detail, transcripts, and methodology → BENCHMARKS.md .

The true story

Same 14 repositories, same 7,213 query/gold pairs as every tool here, CodeGraph included. Full methodology, every raw number, and the other 9 tools →