vinemap 0.1.1

Vinemap — a local-first, graph-based context engine for AI coding agents

Vinemap — Local Code Graph & MCP Context for AI Coding Agents

Vinemap builds a local code graph of your repository and delivers token-budgeted context packs to AI coding agents — Claude Code, Cursor, Codex CLI, Gemini CLI, Copilot, and any MCP client.

Stop wasting tokens on blind exploration. Vinemap ranks the right files before your agent's first tool call.

Install

pip install vinemap

Python 3.9+. Zero runtime dependencies.

Quick start

vinemap index .                              # build code graph → .vinemap/
vinemap query "where is auth handled"        # ranked search
vinemap pack  "how do sessions work" --budget 6000
vinemap mcp .                                # MCP server (stdio)
vinemap connect cursor                       # auto-configure Cursor

MCP integration

{
  "mcpServers": {
    "vinemap": {
      "command": "vinemap",
      "args": ["mcp", "/path/to/project"]
    }
  }
}

Tools: graph_retrieve · graph_read · graph_neighbors · graph_stats

Why Vinemap?

• Graph-native context — files, symbols, imports, call edges (not just text chunks)
• Token-budgeted packs — <codebase_context> blocks sized to your budget
• Session memory — remembers touched files and decisions across turns
• 100% local — nothing leaves your machine
• MCP-native — works with Cursor, Claude Code, Codex, Gemini, and more

Supported languages

Python (AST), TypeScript, JavaScript, Go, Rust, Java, C/C++, C#, Ruby, PHP, Kotlin, Swift.

Links

• GitHub: github.com/rohit0x62/vinemap
• Docs & roadmap: github.com/rohit0x62/vinemap/tree/main/docs
• Website: vinemap.dev

License

Apache-2.0