Vera

Overview

Vera is a high-performance local code search tool written in Rust that combines BM25 keyword matching, vector similarity, and cross-encoder reranking. It supports 65+ programming languages through tree-sitter parsing and runs entirely offline, returning structured results with file paths, line ranges, and symbol metadata. As an MCP server, it exposes search_code, get_stats, get_overview, and regex_search tools, making it suitable for AI agent workflows. The tool achieves impressive benchmark results with MRR@10 of 0.91, outperforming other code search solutions.

Try asking AI

After installing, here are 5 things you can ask your AI assistant:

• search_code

  Searches code across the codebase using BM25, vector similarity, and cross-encoder reranking

  Try:Find authentication logic in the codebaseTry:Search for OAuth token refresh handling
• get_stats

  Returns statistics about the current code index

  Try:Show me index statistics for this projectTry:What's the current size of my Vera index?
• get_overview

  Provides a project overview including call graphs and structure information

  Try:Give me an overview of this project structureTry:What are the main components in this codebase?
• regex_search

  Performs regex pattern matching across the codebase

  Try:Find all functions named 'main' using regexTry:Search for TODO comments in the code

Comparable tools

Installation

bunx @vera-ai/cli install   # or: npx -y @vera-ai/cli install / uvx vera-ai install

vera setup                       # Interactive wizard
vera setup --api                 # API mode (recommended)

vera mcp   # or: bunx @vera-ai/cli mcp / uvx vera-ai mcp

{
  "mcpServers": {
    "vera": {
      "command": "vera",
      "args": ["mcp"]
    }
  }
}

FAQ

What file types does Vera support?

Vera supports 65+ programming languages through tree-sitter parsing, with 61 languages having full tree-sitter grammars compiled in.

Does Vera require an internet connection?

Vera can run completely offline when using local models. It also supports API mode for model inference that requires internet connectivity to an OpenAI-compatible endpoint.

Compare Vera with