Sensegrep
Let agents find code by meaning and structure—not just text matches—across the repo you index with Sensegrep.
Overview
Sensegrep is a MCP server for the Build phase that provides semantic and structural code search tools for AI-native development.
What is this MCP server?
- @sensegrep/mcp npm package version 1.5.3 with stdio MCP transport
- Combines semantic (embedding) search with structural code search for AI-native workflows
- SENSEGREP_ROOT configures index root; defaults to current working directory
- Embeddings providers: gemini, openai, or bedrock via SENSEGREP_PROVIDER
- GEMINI_API_KEY required only when provider is gemini
- Catalog version 1.5.3 for @sensegrep/mcp npm package
- Three documented embedding providers: gemini, openai, bedrock
- Stdio transport with SENSEGREP_ROOT and SENSEGREP_PROVIDER environment configuration
Community signal: 1 GitHub stars.
What problem does it solve?
Agents waste tokens and miss edits because text search alone cannot reliably find code by behavior or architecture.
Who is it for?
Solo builders on medium or large repos who already use MCP agents and can supply Gemini, OpenAI, or Bedrock embeddings.
Skip if: Tiny single-file scripts, teams forbidden from sending code to embedding providers, or builders who only need Markdown note retrieval.
What do I get? / Deliverables
After configuring Sensegrep, your agent can query an indexed codebase with embedding-backed semantic search plus structural queries through MCP.
- Indexed semantic and structural search tools callable from your coding agent
- Faster, more accurate file targeting for refactors and feature work
Recommended MCP Servers
Journey fit
Sensegrep is Build agent-tooling because it augments how coding agents navigate and understand codebases during implementation. Agent-tooling fits semantic plus structural search exposed as MCP tools, distinct from shipping checks or launch SEO work.
How it compares
Codebase semantic search MCP, not a project sprint brain or Lightning-paid AI utility bundle.
Common Questions / FAQ
Who is Sensegrep for?
Sensegrep is for indie developers and agent users who need semantic plus structural search across a local repo root from Claude Code, Cursor, or similar MCP clients.
When should I use Sensegrep?
Use Sensegrep during Build when navigation and refactors depend on finding related code by meaning or structure, not exact string matches alone.
How do I add Sensegrep to my agent?
Add stdio MCP for @sensegrep/mcp 1.5.3, set SENSEGREP_ROOT to your project path, choose SENSEGREP_PROVIDER (gemini, openai, or bedrock), and provide GEMINI_API_KEY when using gemini.