Sibyl-Memory

Durable, file-based long-term memory for AI agents. Five-package plugin family: SDK, CLI, MCP server, Hermes adapter, and a LangGraph BaseStore. No vector datab

What is Sibyl-Memory?

Durable, file-based long-term memory for AI agents. Five-package plugin family: SDK, CLI, MCP server, Hermes adapter, and a LangGraph BaseStore. No vector database, no embeddings.

Overview of Sibyl-Memory

What it is

Sibyl-Memory is a durable, file-based long-term memory infrastructure for AI agents. It is a family of five PyPI packages: sibyl-memory-client (local-first SDK), sibyl-memory-cli (command-line interface), sibyl-memory-mcp (MCP server), sibyl-memory-hermes (Hermes Agent adapter), and sibyl-memory-langgraph (LangGraph BaseStore adapter). The core engine is SQLite-backed with a five-tier hierarchical schema, FTS5 full-text search, and multi-tenant design. No vector database, no embedding model, and no external retrieval service are required. Memory data lives on the agent's machine as a single file on disk. The architecture was benchmarked on LongMemEval Oracle (ICLR 2025) and placed #2 overall at 95.6%, the only file-based system in the top tier.

Key features

  • File-based long-term memory with no vector database or embeddings.
  • Five-package plugin family: SDK, CLI, MCP server, Hermes adapter, LangGraph BaseStore.
  • SQLite-backed, five-tier hierarchical schema, FTS5 search, multi-tenant.
  • Local-first and privacy-preserving: memory content never leaves the machine.
  • Achieved #2 overall (95.6%) on LongMemEval Oracle benchmark (ICLR 2025).
  • Supports MCP-compatible clients (Claude Code, Codex, Cursor, Continue).
  • Self-learning skill detection and local memory linter (available in Pro/Elite tiers).

Use cases

Sibyl-Memory is designed for AI agents that need durable, long-term memory across sessions. It can be used with Hermes Agent (v0.13+), LangGraph (as a BaseStore), or any MCP-compatible client. The file-based architecture makes it suitable for local-first, privacy-sensitive deployments where no external vector database or embedding service is desired.

Getting started

bash pip install sibyl-memory-cli sibyl init sibyl init opens a browser to activate an account (free tier by default). For direct SDK use: pip install sibyl-memory-client. For Hermes integration: pip install sibyl-memory-cli then sibyl setup hermes. For MCP clients: pip install sibyl-memory-mcp sibyl-memory-cli then sibyl setup claude-code (or other client). See the repository README for full details.

License and pricing

The code in this repository is licensed under the MIT License (open source). An account is required to use the memory tools; the free tier is the default. Pro and Elite tiers unlock self-learning, the memory linter, and remove the local storage cap. Pricing details for paid tiers are managed through the account system.

Based on the project's public README and repository metadata.

Add this badge to your README

Featured on YunoDB Badge styles & help

Maintain this project? Show that it is featured on YunoDB.

Markdown[![Featured on YunoDB](https://yunodb.com/badge/sibyl-labssibyl-memory.svg)](https://yunodb.com/tools/sibyl-labssibyl-memory?utm_source=github&utm_medium=badge)
HTML<a href="https://yunodb.com/tools/sibyl-labssibyl-memory?utm_source=github&utm_medium=badge"><img src="https://yunodb.com/badge/sibyl-labssibyl-memory.svg" alt="Featured on YunoDB" height="20"></a>