Sensegrep
What is Sensegrep?
Sensegrep is an open-source semantic code search tool that combines AI embeddings with tree-sitter AST parsing so AI coding agents can find code by meaning rather than by text pattern.
Key Capabilities
- Semantic Search: Matches concepts rather than literal strings, so a query for "authentication logic" surfaces auth functions even when the word does not appear in the code.
- Structural Retrieval: Uses tree-sitter AST parsing for AST-aware, language-accurate matching.
- Exact Matching: Supports literal and regex patterns for occurrence audits and verification after semantic discovery.
- Metadata Filtering: Provides more than 30 symbol-metadata filters to narrow results.
- Context Reduction: Detects logical duplicates and tree-shakes output to collapse irrelevant code, shrinking the context delivered to agents.
- Agent Integration: Ships as an MCP server and a Claude Code plugin, plus a standalone CLI.
Audience & Use Cases
- Audience: AI-assisted developers and agent-pipeline builders.
- Use Case: Routing AI agents to the relevant code during discovery, concept search, and behavioral questions across a codebase.
Technical Specifications
- Language: TypeScript, distributed via npm as @sensegrep/core and @sensegrep/cli.
- License: Apache 2.0.
- Embeddings: Local Ollama embeddings with an optional remote evaluation path for evidence-aware reranking.
Categories & Use Cases
Technical Details
| Mobile Application | No |
|---|
FAQs
What is Sensegrep?
Sensegrep is an open-source semantic code search tool that combines AI embeddings with tree-sitter AST parsing so AI coding agents can find code by meaning rather than by text pattern.