◉
Local by default
SQLite, local embeddings and deterministic fallback. Your context stays under your control.
Local-first · MCP-native · Open source
Your AI and your work finally remember.
LEVH is the local-first memory layer for AI agents and humans. It keeps context alive, signals what matters, and lets stale noise fade.
$ pip install levh
Real local run — 20 demo memories, a persistent entity graph, and one conflict candidate for human review. No cloud account or telemetry required.
A memory system, not a transcript dump
Capture. Understand. Surface the right thing.
LEVH turns scattered workflow context into a local, queryable memory layer your AI tools can actually use.
⌁CaptureMCP, CLI, API or connectors
→
✦UnderstandTrust, entities, conflicts, decay
→
↗RecallRight context, when it matters
⌘
MCP-native
59 tools for capture, recall, graph, trust, review, decisions and more.
↻
Memory that ages
Reinforce useful context. Let stale details fade. Keep important rules pinned.
◇
Real context graph
People, projects, organizations and decisions become connected, inspectable context.
✓
Explainable trust
Reliability signals are deterministic and inspectable—not a black-box truth score.
⌁
One process
Dashboard, REST API, WebSocket and MCP server without Redis or a managed database.
Start in two minutes
Give your agents a memory.
Install locally, verify the runtime, then connect your MCP client.
$ pip install levh && levh doctor