Embgrep
Give your coding agent embedding-powered semantic search over local files so it finds relevant code without exact-match grep or cloud index services.
Overview
Embgrep is an MCP server for the Build phase that provides local embedding-powered semantic search over files with no external services.
What is this MCP server?
- Embedding-powered grep: meaning-based search across local files
- Zero external services in the server description—indexes and search stay on your machine
- embgrep PyPI package (0.1.0) with stdio MCP transport
- Complements literal ripgrep when symbols and wording differ from how you ask
- GitHub source QuartzUnit/embgrep for self-hosted agent setups
- MCP server version 0.1.0
- PyPI identifier embgrep, stdio transport
- Repository QuartzUnit/embgrep on GitHub
What problem does it solve?
Agents waste tokens and time when ripgrep misses the right files because your question uses different words than the code.
Who is it for?
Solo builders who want private, local semantic code search wired into MCP agents alongside their normal editor workflow.
Skip if: Teams that only search logs in production observability stacks, or who need org-wide cross-repo search with centralized team permissions.
What do I get? / Deliverables
After registering embgrep, your agent can semantically search the local tree so you find related implementations faster without a cloud code index.
- Agent-accessible semantic search over configured local paths
- Private search workflow with no required external search service
- Faster navigation hooks for build, review, and fix loops
Recommended MCP Servers
Journey fit
Semantic repo search is primarily adopted while you equip agents and local dev workflows during build, before you rely on it in review and fixes. Agent-tooling is the shelf because embgrep is an MCP stdio server that extends what your agent can query in the workspace, not a deploy or monitoring product.
How it compares
Local semantic file-search MCP server, not a document OCR extractor or RSS research feed kit.
Common Questions / FAQ
Who is embgrep for?
Developers using MCP-enabled agents who want meaning-based code search on their machine without subscribing to external embedding search APIs.
When should I use embgrep?
Use it when you are navigating or refactoring a codebase and need to locate concepts, patterns, or related modules that literal grep will not catch.
How do I add embgrep to my agent?
Install embgrep from PyPI (0.1.0), add it as an MCP stdio server with identifier embgrep in your host JSON, restart, and invoke its search tools from the agent.