Skip to content
MCP server

haiku.rag

Local-first RAG that answers from your documents with page and section citations

Runs locallyStatus and history
Pricing
Open source
Platforms
CLI, self-hosted, Linux
Runs
Locally
Version
0.112.0
Listing updated
3 days ago

About haiku.rag

haiku.rag is an open-source MCP server for knowledge and memory. It runs locally on your machine and is started by your AI client.

Answers can be checked: each carries the page number and section heading it came from, and visual grounding highlights the cited chunk on the original page image. An optional capability makes every answer declare what grounds it, including saying that nothing does.

It indexes PDFs, office documents, HTML, Markdown and images through Docling, retrieves with hybrid vector and full-text search plus optional reranking, and can search several databases at once. For questions that need counting or aggregation, the agent runs sandboxed Python across the documents. Storage is an embedded LanceDB, and by default models come from Ollama so no server or API key is required; S3, GCS, Azure and LanceDB Cloud also work.

Besides the MCP server for Claude Code, Codex and Claude Desktop it offers a Python API, Pydantic AI capabilities, a terminal chat, a web app and a haiku-ingester service for continuous ingestion. Requires Python 3.12 or newer and is MIT licensed.

Install

Run in a terminal

claude mcp add haiku-rag -- uvx haiku-rag

Add to claude_desktop_config.json

{
  "mcpServers": {
    "haiku-rag": {
      "command": "uvx",
      "args": [
        "haiku-rag"
      ]
    }
  }
}
Install in CursorOpens Cursor, which asks you to confirm. Or add it by hand:

Add to ~/.cursor/mcp.json

{
  "mcpServers": {
    "haiku-rag": {
      "command": "uvx",
      "args": [
        "haiku-rag"
      ]
    }
  }
}
Install in VS CodeOpens VS Code, which asks you to confirm. Or add it by hand:

Add to .vscode/mcp.json

{
  "servers": {
    "haiku-rag": {
      "type": "stdio",
      "command": "uvx",
      "args": [
        "haiku-rag"
      ]
    }
  }
}

Add to ~/.codeium/windsurf/mcp_config.json

{
  "mcpServers": {
    "haiku-rag": {
      "command": "uvx",
      "args": [
        "haiku-rag"
      ]
    }
  }
}

Add to your client's MCP settings (most use this shape)

{
  "mcpServers": {
    "haiku-rag": {
      "command": "uvx",
      "args": [
        "haiku-rag"
      ]
    }
  }
}

What it does

Cited answers
Page numbers and section headings accompany answers.
Visual grounding
Highlights the cited chunk on the page image.
Hybrid retrieval
Vector and full-text search with optional reranking.
Sandboxed analysis
Runs Python across documents for computation.
Continuous ingestion
Filesystem, HTTP, S3 and WebDAV sources with a job queue.

Good for

  • Ask questions of a folder of PDFs with sources
  • Give Claude Desktop a private document search
  • Check an answer by viewing the cited page
  • Build a Pydantic AI agent on your own corpus

Pricing

Open source

MIT licensed

Prices change. Check ggozad.github.io before you buy.

haiku.rag FAQ

Does it need an API key?

No by default; it uses Ollama and an embedded LanceDB database.

What file types?

PDFs, office documents, HTML, Markdown and images.

Which Python version?

Python 3.12 or newer.

Is haiku.rag free?

Yes. haiku.rag is open source, so you can use and self-host it for free.

How do I add haiku.rag to Claude Code?

Run this command in your terminal: claude mcp add haiku-rag -- uvx haiku-rag

What are the best alternatives to haiku.rag?

The closest alternatives to haiku.rag are Codebase Memory, XERJ, Claude Code Ultimate Guide, Nautilus Compass Memory and note1.

Alternatives to haiku.rag

  1. Codebase Memory

    MCP server

    Local MCP server that builds a code knowledge graph for AI coding agents

    Runs locally
    • Open source
    • 48.7K GitHub stars
    • Knowledge and memory
  2. XERJ

    MCP server

    Local search engine for AI agents with cited answers over your code and docs

    Runs locally
    • Open source
    • 19.2K GitHub stars
    • Knowledge and memory
  3. MCP server for searching the Claude Code Ultimate Guide from any MCP client

    Runs locally
    • Open source
    • 6.5K GitHub stars
    • AI models and LLM tooling
  4. Local-first long-term memory layer for AI agents, served over MCP

    Not checked yet
    • 5 tools
    • Remote
    • Open source
    • 1.3K GitHub stars
    • Knowledge and memory
  5. note1

    MCP server

    MCP server giving AI assistants search and summaries across your note1 meetings

    • 4 tools
    • Remote
    • Free trial
    • 1 GitHub star
    • Productivity
  6. ASTra

    MCP server

    Let agents read, edit, validate and undo Word .docx files with exact operations

    Runs locally
    • 5 tools
    • Freemium
    • Files and documents