Quick Start

Get vibe-hnindex running in 5 minutes. This guide assumes you have Node.js, Ollama, and Qdrant installed. If not, see the Installation guide first.

1. Install the CLI Helper

The easiest way to set up is with the CLI:

npm install -g hnindex-cli

2. Initialize MCP Config

Pick your editor and run the init command:

Use --cwd /path/to/project for project-based targets. Add flags for custom config:

hnindex init --mcp antigravity \
  --ollama-url http://localhost:11434 \
  --ollama-model bge-m3:567m \
  --qdrant-url http://localhost:6333

3. Or Add Config Manually

If you prefer manual configuration, add this JSON to your MCP settings file:

{
  "mcpServers": {
    "vibe-hnindex": {
      "command": "npx",
      "args": ["-y", "vibe-hnindex"],
      "env": {
        "OLLAMA_URL": "http://localhost:11434",
        "OLLAMA_MODEL": "bge-m3:567m",
        "QDRANT_URL": "http://localhost:6333"
      }
    }
  }
}

4. Install the Agent Skill (Recommended)

The vibe-hnindex skill teaches your AI assistant how to use all 20+ tools effectively — search modes, streaming, fuzzy matching, benchmarks, and best practices. Without the skill, the AI may not know about advanced features like stream: true, fuzzy search, or Code Agent.

hnindex init-skill --target claude          # Claude Code
hnindex init-skill --target antigravity     # Google Antigravity
hnindex init-skill --target cursor          # Cursor
hnindex init-skill --target windsurf        # Windsurf
hnindex init-skill --target codex           # OpenAI Codex
hnindex init-skill --target vscode          # VS Code
hnindex init-skill --target openclaw        # OpenClaw

This creates a SKILL.mdfile in your editor's skills directory that the AI automatically loads on startup.

5. Restart Your AI Tool

Restart Claude, Cursor, Antigravity, or whatever MCP client you use.

6. Index Your First Project

In your AI chat, type:

Index the codebase at /path/to/my-project, name it my-project

This calls the index_codebase tool. The first index may take a while (embeddings are being generated), but subsequent runs are incremental.

7. Search Your Code

Now you can search:

Search my-project for authentication middleware
Search my-project for "error handling" in mode: semantic
Search my-project for "/TODO|FIXME/g" in mode: regex
List all indexed projects

8. Try Smart Context

For complex tasks, use smart context to gather relevant code automatically:

Get smart context for my-project, task: add rate limiting to the API

What's Next?

• Configure environment variables for custom behavior
• Learn about all search modes — keyword, semantic, hybrid, regex, symbol
• Explore indexing options — parallel workers, ignore files, watch mode
• Use Smart Context for AI-assisted code understanding