Without CodeIR
⏺Searched for"session" "cached"
╸ 23 matches across 14 files
⏺Readsessions/backends/cached_db.py
╸ reading 135 lines
⏺Searched for"SessionStore" "save"
╸ 9 matches across 5 files
⏺Readsessions/backends/db.py
╸ reading 118 lines
⏺Searched for"cache_key" "expiry"
╸ 31 matches across 11 files
⏺Readsessions/base_session.py
╸ reading 53 lines
still searching▊
With CodeIR
⏺Read.claude/bearings.md
╸ codebase mapped — 247 modules
⏺codeir searchcached_db session
SAVE.11SessionStore.savecached_db.py:88[method]
ASV.03SessionStore.asavecached_db.py:95[async]
… +12 more (ctrl+o to expand)
⏺codeir showSAVE.11--level L1
Entity:SessionStore.save [method]
File:cached_db.py:88
… +2 lines (ctrl+o to expand)
⏺codeir impactSAVE.11--depth 2
Affected: 12 entities across 5 files
… +8 lines (ctrl+o to expand)
⏺codeir expandSAVE.11
File:cached_db.py:88–93
… +6 lines (ctrl+o to expand)
⏺ Ready to implement
Give Claude a map of your codebase — not just a search bar.
Compile any Python repo into a navigable semantic index. Understand first, then code.
pip install codeir-tools
Don't take our word for it. Take theirs.
This changes how I navigate unfamiliar codebases entirely.
Claude Opus
I found the bug in six queries across 38,000 entities. I'd still be grepping.
Claude Sonnet
Wait, why is this not sitting under every model?
GPT-5.3
This solves the "Stochastic Parrots" problem.
Gemini
CodeIR occupies a unique niche: deterministic, structural, LLM-optimized code understanding.
DeepSeek
Ask your model about CodeIR.
Paste How It Works into any frontier model. See what it thinks.