Show HN: Sqry – semantic code search using AST and call graphs
15 hours ago
- #code-search
- #security
- #developer-tools
- Text search limitations: misses structure, can't show callers, returns, or usage.
- Embeddings are slow (~1,400ms) and imprecise (~60% precision).
- Cloud tools require uploading code, not suitable for private/regulated code.
- sqry offers structural queries with exact answers, parsing code like a compiler.
- MCP Integration allows AI assistants to get exact structural answers.
- Features include semantic_search, direct_callers, call_hierarchy, find_unused, semantic_diff.
- LSP server support with 27 custom methods, works in editors.
- Supports 35 languages with full relation tracking in 28 languages.
- Fast indexing: Linux kernel (28M lines) indexed in under 2 minutes.
- Natural language queries via 'sqry ask' with 99.75% intent accuracy.
- Security features: signed builds, SBOMs, audited dependencies, continuous fuzzing.
- Mutation testing, coverage gates, unsafe code tracking, and code signing policy.
- Privacy policy ensures no data transfer unless requested.
- Open-source with support subscriptions for enterprise support.