VonderVuflya/Yggdrasil

Overview

Yggdrasil provides durable, local-first memory for AI coding agents via the Model Context Protocol (MCP). It stores data in SQLite and Markdown, with zero dependencies, curated content, and semantic deduplication, giving users full ownership of their data. It integrates with Claude Code, Codex, and any MCP host.

Best for

Best for
Developers who want a lightweight, local memory layer for MCP-compatible AI coding agents

Use cases

• Persisting agent context across coding sessions without cloud storage
• Building a curated, deduplicated knowledge base for AI assistants
• Running memory-backed agents locally with full data control

How to use

Tools exposed

• all-minilm
• nomic-embed-text
• mxbai-embed-large
• paraphrase-multilingual
• bge-m3

Tested with

Claude Desktop, Claude Code, Cursor, ChatGPT

Notes

Yggdrasil provides durable, local-first memory for AI coding agents via the Model Context Protocol (MCP). It stores data in SQLite and Markdown, with zero dependencies, curated content, and semantic deduplication, giving users full ownership of their data. It integrates with Claude Code, Codex, and any MCP host.

15 stars on GitHub. Last updated 2026-07-04. Licensed AGPL-3.0.

Use cases

• Persisting agent context across coding sessions without cloud storage
• Building a curated, deduplicated knowledge base for AI assistants
• Running memory-backed agents locally with full data control

Pros

• Zero external dependencies simplifies setup and maintenance
• Local-first design ensures data privacy and ownership
• Semantic deduplication reduces clutter in stored memories

Cons

• Small community (15 stars) may mean limited support and updates
• Python-only implementation restricts use in non-Python environments
• Requires MCP host compatibility, limiting standalone utility

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