Levh

Levh is an open-source, local-first memory layer for AI agents and the people who work alongside them. It captures context from conversations, projects, files, email, calendars, and MCP tools, then builds an inspectable knowledge graph that links people, projects, organizations, and decisions into relationships you can read and audit. Rather than a black-box score, Levh surfaces deterministic trust and conflict signals so you can see why a piece of context is considered reliable, and it lets stale details fade over time while important guidance stays pinned. Everything runs on your own machine: it stores data in SQLite with local embeddings and a deterministic fallback, and needs no cloud account, no managed database, and no Redis. Levh is MCP-native, exposing 59 tools that cover capture, recall, graph navigation, trust assessment, review, and decision support, and it also ships a dashboard, a REST API, and a WebSocket interface alongside the MCP server. The workflow is capture, understand, then recall: context comes in through MCP, CLI, API, or connectors; Levh analyzes trust, entities, conflicts, and decay; and it resurfaces the right context when an agent or person needs it. It installs with pip install levh and is released under AGPL-3.0. Levh suits developers and teams who want persistent agent memory they can inspect and keep on their own hardware, instead of a hosted memory service.

Best for developers and teams who want persistent, inspectable AI agent memory that runs locally with no cloud account.