Grepai, a free MCP server tool by Yoanbernabeu, is built in Go and designed to augment AI agents' capabilities within codebases. It provides semantic context, enabling intelligent interactions and reducing token usage for agents like Claude Code and Cursor. Grepai operates locally, prioritizing privacy, and supports various embedding providers to enable advanced code understanding for developers and AI alike. Its architecture allows for efficient real-time indexing and search.

What does Grepai connect to?

Grepai functions as an MCP server tool, connecting to AI agents such as Claude Code, Cursor, and Windsurf. It supplies these agents with semantic context derived from a codebase. For its core operation, Grepai requires an embedding provider. Ollama is the recommended local, privacy-first option, ensuring data remains on-premises. Additionally, it supports external embedding providers like LM Studio and OpenAI, offering flexibility in how code embeddings are generated and managed. This setup allows AI agents to gain a deeper understanding of project specifics.

Core features and tools

Grepai exposes several callable tools for code interaction. These include grepai init for project setup and grepai watch to start an indexing daemon that scans, embeds, and monitors file changes. Semantic code searches are performed with grepai search "query" using natural language. For understanding code flow, grepai trace callers "function" identifies calling functions, and grepai trace callees "function" finds called functions. Indexing status is checked via grepai status. Additionally, grepai mcp-serve tests the server, and grepai agent-setup helps configure non-MCP agents. Workspace management is handled by commands like grepai workspace create <name> and grepai workspace list.

Who is it for?

Grepai is intended for developers working with large codebases and for enhancing AI coding agents. It helps both human users and AI agents understand code more deeply by providing semantic context. Developers can use it to navigate complex projects, while AI agents benefit from reduced input tokens and improved code interaction. This tool is particularly useful for those seeking to integrate AI into their development workflows for tasks like code search and dependency tracing. Skip this if your development workflow does not involve AI agents or semantic code understanding.

How to get started

No credentials are required for Grepai's core functionality when using local embedding providers. If an external embedding provider like OpenAI is chosen, an OpenAI API key is needed for that specific service, which is communicated through Grepai's Native MCP server or JSON output.

Use cases

• Semantic Code Search: Query a codebase using natural language (e.g., "authentication logic") to find relevant code segments based on their meaning.
• Call Graph Tracing: Trace function dependencies (e.g., "grepai trace callers Login") to understand the impact of potential code changes before implementation.
• AI Agent Context Provisioning: Act as an MCP server to supply AI agents with a semantic map of the codebase, reducing input tokens and enabling intelligent code interaction.

Grepai provides a local, privacy-first solution for integrating AI agents with codebases, offering semantic context and advanced search capabilities. Its Go-based architecture supports real-time indexing and various embedding providers. It helps developers and AI agents navigate complex projects efficiently, reducing the token burden on AI models. This tool is valuable for enhancing code understanding and accelerating development workflows.