EverOS

EverOS is a Python library and memory runtime for developers building agents that need context across sessions and applications. It keeps conversations, files and task histories in readable Markdown that you can edit and version with Git.

• Persistent memory: Store user preferences and past interactions separately from agent task records, with retrieval scoped to users, agents, applications, projects or sessions.
• Search: Retrieve memories through keyword or hybrid keyword/vector search; hybrid retrieval requires an embedding provider, and agentic search also requires reranking.
• Reusable skills: Consolidate past interactions and turn successful task histories into reusable procedures; reflection and skill extraction require an embedding provider.
• Files and knowledge pages: Ingest images, PDFs, audio and office files with a multimodal provider and parser support, or maintain editable Knowledge Wiki pages linked to source material with embedding and reranking providers.

EverOS integrates with Claude Code, OpenClaw and MCP, and is compatible with Codex, Hermes, and the OpenAI and Anthropic SDKs. Self-hosted deployments keep Markdown and SQLite/LanceDB indexes on your hardware. Memory processing requires an LLM provider, so local storage alone doesn't guarantee offline operation or keep provider requests on your machine.

The software is open source under Apache 2.0. EverOS Cloud provides managed hosting with the same memory format and retrieval engine, and Markdown export lets you move memories between deployments.