Lorekeeper

Name: Lorekeeper
Author: Jessinra

@Jessinra

リポジトリホームページ

Lorekeeper について

Self-improving memory for AI agents

#ai-agent #memory #ai #agentic-ai #codex #claude-code #cursor

基本情報

カテゴリ

その他

ライセンス

Apache-2.0

ランタイム

html

トランスポート

stdio

公開者

Jessinra

投稿者

Jessin Ra

設定

以下の設定を使って、このサーバーを MCP 対応クライアントに追加してください。

{
  "mcpServers": {
    "lorekeeper": {
      "command": "lorekeeper"
    }
  }
}

ツール

ツールは検出されませんでした

ツールは README から自動的に抽出されます。メンテナーは ## Tools という見出しの下に記載することで、このタブに反映できます。

概要

What is Lorekeeper?

Lorekeeper is a self-improving memory server for AI agents. It stores memories locally in SQLite on your own disk, connects to any MCP-compatible agent, and improves with every session. It is designed for solo developers and agent workflows that need zero ops, zero cloud, and a persistent memory store.

How to use Lorekeeper?

Install with pip install lorekeeper-mcp, then run lorekeeper setup to configure your agents, and start the server with lorekeeper. Use lorekeeper-dashboard to open a web UI for managing memories. Agents can then call tools like lore_remember and lore_search to store and retrieve memories.

Key features of Lorekeeper

Hybrid semantic + keyword search with time-decay and usage scores
Self-improving quality loop: feedback adjusts memory scores
Auto-linking: new memories linked to closest semantic neighbors
Duplicate detection: blocks near-identical content
Full web dashboard with seven tabs for management
Universal MCP support: works with many agents
Local-first: data stays on your machine, no cloud
Namespaces: multiple agents share one store isolated
Reflection: agents auto-extract learnings from sessions

Use cases of Lorekeeper

Staying in context across sessions: agents remember preferences and code patterns
One memory pool, multiple agents: Claude Code, Cursor, Hermes share knowledge
Cross-session debugging: recall root causes and fixes from weeks ago
Project onboarding: new agents start with architecture context from past sessions

FAQ from Lorekeeper

What makes Lorekeeper different from file-based memory or cloud services?

Lorekeeper is a local MCP server that uses hybrid search, self-improving memory scores, and auto-linking. It is a single pip install with no cloud dependency, no API keys, and no configuration. File-based tools require manual maintenance, and cloud services charge per API call and send data off your machine.

How do I install and set up Lorekeeper?

Run pip install lorekeeper-mcp, then lorekeeper setup to auto-detect and configure supported agents. Finally, start the server with lorekeeper. No additional configuration is needed.