Sibyl-Memory

Name: Sibyl-Memory
Rating: 86 (1 reviews)
Author: Sibyl-Labs

MCP Server

Sibyl-Labs/Sibyl-Memory

Sibyl-Memory is a local-first, SQLite-backed agentic memory infrastructure that uses a five-tier hierarchical schema and FTS5 search without relying on vector databases or embedding models.

99 10MITUpdated 2026-08-13

Great

View on GitHub Open source file

Overview

Sibyl-Memory provides a complete ecosystem of PyPI packages designed to deliver durable, structured, and multi-tenant memory for AI agents. Featuring a strict five-tier hierarchical model (Hot, Warm, Cold, Reference, Archive) and SQLite-backed full-text search, it eliminates external embedding dependencies while achieving high performance benchmarks. The suite includes an SDK, CLI for tier management, a Model Context Protocol (MCP) server, a LangGraph store adapter, and a Hermes Agent integration.

Capabilities

▸Five-tier hierarchical memory schema
▸SQLite-backed full-text search (FTS5)
▸Multi-tenant storage architecture
▸Model Context Protocol (MCP) server support
▸LangGraph BaseStore integration
▸Local-first operation with optional tiered verification

Best for

Providing long-term memory and context retention across sessions for AI coding assistants like Claude Code and Cursor, Integrating durable agentic storage into LangGraph workflows via a custom BaseStore adapter, Managing local-first, privacy-preserving agent state and entity ledgers without incurring vector database or embedding costs

Works with

Claude CodeCursorCodex

Evoa Score breakdown= Σ (score × weight)

Task usefulness20%100 → +20.0

実タスクにどれだけ役立つか（機能の豊富さ・用途の明確さ）。 — AIによるcapabilities/use-cases解析

Code quality15%90 → +13.5

実装・指示の品質。 — AIによるSKILL.md/README解析

Maintenance15%100 → +15.0

リポジトリがどれだけ活発に保守されているか。 — GitHub 最終push日時の新しさ

Documentation12%47 → +5.6

ドキュメントの充実度・分かりやすさ。 — README/独自要約の情報量

Security15%100 → +15.0

危険・不審な挙動が無いか。 — AIによるセキュリティレビュー

Originality10%85 → +8.5

ありふれたラッパーではない独自性。 — AIによる独自性判定

Popularity8%52 → +4.2

コミュニティの採用度。 — GitHub Stars/Forks（対数スケール）

Compatibility5%75 → +3.8

対応AIエージェントの広さ。 — AIによる対応エージェント判定

Weighted total85.5 / 100

ライセンス不明/制限あり（Red）のSkillは総合スコアに0.85倍の補正を適用します。ランキングはこのScoreのみで決まり、広告で変わりません。算出方法の詳細 →

Security considerations

Memory contents remain completely local on the user's machine. The only outbound network calls occur during optional tier/subscription verification, which transmits account metadata and byte size metrics only.