Semantic search that finds what code does, not just what it says. Works locally. Scales across repos. Free for production use.
Source-available · BSL 1.1 · 18 languages · Works offline
42 repos indexed for $0.70
Keyword matches only, no cross-repo awareness. Searching "error handling" returns every comment and string literal — not actual error handlers.
One project at a time. Can't show patterns across your architecture. No structural awareness beyond the current workspace.
GitHub-only, no cross-repo intelligence, no semantic understanding. And it costs money at scale.
Your teams search code every day. They shouldn't need three tools.
Finds code intent, not strings.
Search "authentication middleware" — grep returns noise, codeindex returns validators, handlers, and schemas.
See dependencies across your architecture.
`codeindex xref UserDTO` finds every caller across all indexed repos. Map blast radius in seconds, not hours.
Claude, Cursor, Windsurf integrate instantly.
Session caching, batch queries, multi-agent deduplication. Your AI tools get smarter context with less cost.
330 queries · 19 repos · 16 languages · vs expert grep patterns
Tested on kubernetes, elasticsearch, langchain, discourse, and 18 more public repos
Search "payment handling" and find all payment logic in seconds — no grepping through directory trees.
Session caching (20-30% savings), batch queries (15-20%), deduplication (10-25%). 42 repos indexed for $0.70.
`codeindex xref` traces dependencies across repos. Map blast radius before you ship.
Local embeddings, pre-embedding secret scanning, scoped tokens, air-gapped deployment. Your code never leaves your network.
Minimal config. BYO API keys or run locally with Ollama.
git clone https://github.com/smart-knowledge-systems/codeindex.git && cd codeindex && bun install && bun src/index.ts initbun src/index.ts search "authentication middleware"| Feature | codeindex | grep | GitHub | Sourcegraph |
|---|---|---|---|---|
| Semantic search | ✓ | ✗ | Limited | ✓ |
| Cross-repo | ✓ | ✗ | ✗ | Config required |
| Self-hosted | ✓ | N/A | ✗ | Partial |
| Agent integration (MCP) | ✓ | ✗ | ✗ | ✗ |
| Source-available | ✓ | ✓ | ✗ | ✗ |
| Cost | Free + infra | Free | $$ | $$$ |
We tested this. Grep wins in specific scenarios. We're honest about tradeoffs.
grep is fine — the overhead isn't worth it.
Use `find` or `fd`. We're not a file finder.
Embedding has startup cost. Warm queries are fast, but first run takes time.
We support 18 languages. Adding more is straightforward via tree-sitter, but we won't pretend we cover everything yet.
codeindex is free and self-serve. But if your team needs hands-on training or a purpose-built context layer, we can help.
Teach your team to use coding agents effectively
A 1-2 day hands-on workshop where we train your engineering team on your actual codebase. Your team bought AI coding tools — we make sure they get real value from them.
Not right for: Teams under 5 engineers, or teams already getting strong results from AI coding agents.
We build the context layer and prove it works
We build a structured context layer for your codebase using codeindex, integrate it with your AI tools via MCP, and run a before/after eval on your actual tasks. You see the measured improvement before you pay.
We define success criteria together before starting. If we don't hit them, no charge. You keep everything we build.
Not right for: Codebases under 50K lines of code, or teams that don't use AI coding agents. For smaller projects, codeindex self-serve is likely all you need.
CodeReady Retainer — After a successful Pilot, ongoing context maintenance with automated drift detection, quarterly reviews, and priority support. Starting at $3,000/month.
Built by a knowledge management consultant (MBA, PMP) who builds production software with coding agents and structured context.
I built cidx-cloud — auth, billing, managed MCP server, Cloudflare deployment — in 3 days as a solo developer. That's not a client case study. It's a data point about what's possible with the right approach.
git clone https://github.com/smart-knowledge-systems/codeindex.git && cd codeindex && bun install && bun src/index.ts init