Yggdrasil

Name: Yggdrasil
Author: vondervuflya

@vondervuflya

关于 Yggdrasil

Durable, local-first memory for AI coding agents over MCP — zero-dependency, curated & semantically de-duped, you own the data (SQLite + Markdown). Works with Claude Code, Codex & any MCP host.

#cli #memory #sqlite #mcp #embeddings #developer-tools #codex #ai-agents #claude #rag #local-first #coding-assistant #llm #ai-memory #model-context-protocol #claude-code

基本信息

分类

其他

传输方式

stdio

发布者

vondervuflya

提交者

Sergey Cherkashin

配置

使用下面的配置,将此服务器添加到你的 MCP 客户端。

{
  "mcpServers": {
    "yggdrasil": {
      "command": "uvx",
      "args": [
        "--from",
        "yggdrasil-memory",
        "ygg",
        "mcp"
      ],
      "env": {
        "YGG_ENGINE_URL": "http://127.0.0.1:42069",
        "YGG_ENGINE_TOKEN": "YOUR_TOKEN"
      }
    }
  }
}

工具

未检测到工具

工具是从 README 中自动提取的。维护者可以在 ## Tools 标题下列出工具,即可填充这部分内容。

概览

What is Yggdrasil?

Yggdrasil is a local, persistent memory layer for AI coding agents like Claude Code, Codex, and any MCP-compatible tool. It stores project context, decisions, and gotchas across sessions and projects, eliminating the need to re-explain your work every time a new chat starts. The system runs as a tiny always-on daemon (~21 MB RAM) with zero dependencies and keeps everything on your machine.

How to use Yggdrasil?

Install inside Claude Code with two commands (/plugin marketplace add VonderVuflya/Yggdrasil then /plugin install yggdrasil) or via CLI using uvx --from yggdrasil-memory ygg install, npx yggdrasil-memory install, pipx, pip, Homebrew (brew install VonderVuflya/tap/yggdrasil), or a .mcpb file for Claude Desktop. Once running, agents interact with it through MCP tools: ygg_health, ygg_bootstrap, ygg_search, ygg_recall, ygg_remember, and ygg_materialize. Hooks automatically inject a project status summary (~300 tokens) at the start of each session.

Key features of Yggdrasil

Persistent memory survives across AI sessions and projects
One brain shared by Claude Code, Codex, and any MCP host
Cross-project recall (“this looks like project B — reuse it?”)
Curated, not captured: agent saves only what matters; governance dedupes and archives
Opt-in self-maintenance using a small local model (zero API tokens)
100% local, no cloud, no account, no telemetry

Use cases of Yggdrasil

Stop re-explaining your project’s decisions and gotchas to every new chat
Share the same memory across different AI tools (Claude Code, Codex)
Recall patterns from past projects to suggest reuse in a current one
Run a background consolidation agent that keeps memory tidy without external API calls
Keep all memory as plain Markdown notes viewable in Obsidian

FAQ from Yggdrasil

How does Yggdrasil store memory?

It uses SQLite with FTS5 indexing; no external database is needed. Optionally, local embeddings via Ollama improve recall@1 from 0.77 to 0.93 (recall@3 reaches 1.00). Each memory is also a plain Markdown file accessible in any Markdown editor.

What dependencies does Yggdrasil require?

Zero dependencies out of the box. The tool runs on SQLite + FTS5 bundled with the daemon. If you want enhanced recall, you can optionally connect to a local Ollama instance for embeddings.

Where does my data live?

All data stays completely on your machine. There is no cloud component, no account, and no telemetry. The daemon writes to a local SQLite database.

Which MCP tools does Yggdrasil expose?

It provides ygg_health, ygg_bootstrap, ygg_search, ygg_recall, ygg_remember, and ygg_materialize. Hooks automatically inject a context summary on session start.