Skip to content
MCP server

Kybase

Self-hosted Markdown memory for AI agents with notes you can read and edit

Runs locallyStatus and history
Install
Pricing
Open source
Platforms
self-hosted, CLI, Web
Runs
Locally
Version
0.2.4
Listing updated
2 days ago

About Kybase

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

Kybase gives an agent a persistent place to write what it learns, in plain Markdown stored in Postgres, while you read and correct the same notes in a browser. The site's example has an agent finding an architecture decision record and answering from it, with the section it came from.

The agent can open one section of a long note and append to it without rewriting the rest. Notes link with wikilinks and appear in a graph, every edit keeps the previous version labelled with the app that made it, and search combines keyword and semantic matching. Embeddings can use local Ollama, Google or OpenAI; without any provider, full-text search still works.

There are two ways to run it: npx kybase-mcp for a single local agent over stdio, or a Docker Compose stack for several agents and the web app, reached over HTTPS with OAuth sign-in. It is open source under AGPL-3.0 and has no vendor account or cloud service in between.

Install

Run in a terminal

claude mcp add kybase --env GOOGLE_API_KEY='<GOOGLE_API_KEY>' --env OPENAI_API_KEY='<OPENAI_API_KEY>' --env DATABASE_URL='<DATABASE_URL>' -- npx -y kybase-mcp

Add to claude_desktop_config.json

{
  "mcpServers": {
    "kybase": {
      "command": "npx",
      "args": [
        "-y",
        "kybase-mcp"
      ],
      "env": {
        "GOOGLE_API_KEY": "<GOOGLE_API_KEY>",
        "OPENAI_API_KEY": "<OPENAI_API_KEY>",
        "DATABASE_URL": "<DATABASE_URL>"
      }
    }
  }
}

Add to ~/.cursor/mcp.json

{
  "mcpServers": {
    "kybase": {
      "command": "npx",
      "args": [
        "-y",
        "kybase-mcp"
      ],
      "env": {
        "GOOGLE_API_KEY": "<GOOGLE_API_KEY>",
        "OPENAI_API_KEY": "<OPENAI_API_KEY>",
        "DATABASE_URL": "<DATABASE_URL>"
      }
    }
  }
}

Add to .vscode/mcp.json

{
  "servers": {
    "kybase": {
      "type": "stdio",
      "command": "npx",
      "args": [
        "-y",
        "kybase-mcp"
      ],
      "env": {
        "GOOGLE_API_KEY": "<GOOGLE_API_KEY>",
        "OPENAI_API_KEY": "<OPENAI_API_KEY>",
        "DATABASE_URL": "<DATABASE_URL>"
      }
    }
  }
}

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

{
  "mcpServers": {
    "kybase": {
      "command": "npx",
      "args": [
        "-y",
        "kybase-mcp"
      ],
      "env": {
        "GOOGLE_API_KEY": "<GOOGLE_API_KEY>",
        "OPENAI_API_KEY": "<OPENAI_API_KEY>",
        "DATABASE_URL": "<DATABASE_URL>"
      }
    }
  }
}

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

{
  "mcpServers": {
    "kybase": {
      "command": "npx",
      "args": [
        "-y",
        "kybase-mcp"
      ],
      "env": {
        "GOOGLE_API_KEY": "<GOOGLE_API_KEY>",
        "OPENAI_API_KEY": "<OPENAI_API_KEY>",
        "DATABASE_URL": "<DATABASE_URL>"
      }
    }
  }
}

Configuration

  • KYBASE_DATA_DIRWhere the embedded database and instance secret live. Defaults to ~/.kybase
  • EMBEDDING_PROVIDEREmbedding provider for semantic search
  • OLLAMA_URLOllama endpoint for local embeddings. Without any provider, full-text search still works
  • OLLAMA_MODELOllama embedding model
  • GOOGLE_API_KEYGoogle API key, when EMBEDDING_PROVIDER=google. Sends note text to Google
  • OPENAI_API_KEYOpenAI API key, when EMBEDDING_PROVIDER=openai. Sends note text to OpenAI
  • DATABASE_URLUse an existing Kybase Postgres instead of the embedded one (migrations are left to the app)

Tools this server provides (1)

Taken from the project documentation; the running server may differ.

search_notes
Search notes by keyword and meaning.

What it does

Plain Markdown notes
Everything the agent saves is readable and editable text.
Hybrid search
Finds notes by exact words and by meaning.
Section-level edits
Agents read and append to parts of long notes.
Version history
Each edit keeps the earlier version, labelled by app.
Wikilinks and graph
Notes link together and show as a graph.
Choice of embeddings
Ollama, Google or OpenAI, or full-text only.

Good for

  • Give Claude or Cursor a shared memory across sessions
  • Keep architecture decisions where agents can cite them
  • Review and fix what an agent wrote down
  • Export notes as a Markdown vault zip

Pricing

Open source

AGPL-3.0, free to self-host

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

Kybase FAQ

Is there a Kybase account?

No. The site says it runs on your server or laptop with no Kybase account or cloud service.

What licence is it?

AGPL-3.0.

Do notes leave my machine?

Not with Ollama embeddings; Google and OpenAI embedding providers send text to those services.

Can I get my notes out?

Yes, kybase-mcp export vault.zip writes them as Markdown.

Is Kybase free?

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

How do I add Kybase to Claude Code?

Run this command in your terminal: claude mcp add kybase --env GOOGLE_API_KEY='<GOOGLE_API_KEY>' --env OPENAI_API_KEY='<OPENAI_API_KEY>' --env DATABASE_URL='<DATABASE_URL>' -- npx -y kybase-mcp

What are the best alternatives to Kybase?

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

Alternatives to Kybase

  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