Kyrzin/kybase

Overview

Self-hosted persistent memory for AI agents. Stores memory as Markdown for human readability and exposes it to agents via the Model Context Protocol (MCP). Written in TypeScript.

Best for

Best for
Developers who want a self-hosted, human-readable memory layer for AI agents.

Use cases

• Give AI agents long-term memory across sessions
• Store agent knowledge in editable Markdown files
• Expose memory to agents through an MCP server

How to use

Tested with

Claude Desktop, Claude Code, Cursor, Windsurf, ChatGPT

Example client config

{\n  "mcpServers": {\n    "kybase": {\n      "command": "npx",\n      "args": ["-y", "kybase-mcp"]\n    }\n  }\n}

Notes

Self-hosted persistent memory for AI agents. Stores memory as Markdown for human readability and exposes it to agents via the Model Context Protocol (MCP). Written in TypeScript.

4 stars on GitHub. Last updated 2026-10-02. Licensed AGPL-3.0.

Use cases

• Give AI agents long-term memory across sessions
• Store agent knowledge in editable Markdown files
• Expose memory to agents through an MCP server

Pros

• Self-hosted, so data stays under your control
• Markdown format is transparent and human-editable
• Uses the standard MCP interface for agent integration

Cons

• Requires self-hosting and ongoing maintenance
• Small community with only 4 GitHub stars
• MCP ecosystem is still evolving, so compatibility may shift

Indexed from awesome-mcp-servers-punkpeye and enriched against its public facts.