Awareness

Awareness is a persistent memory cloud for AI agents and copilots. It is meant for cases where an AI tool needs to retain conversations, decisions, and project context across sessions, so it can resume where it left off without being prompted to remember.

Its memory system is described as having five layers: Identity Profiles, Memory Graph, Hybrid Retrieval, Smart Extractors, and Tool Connectors. The Identity Profiles layer is used for granular control over AI behavior, coding styles, and project constraints. The Memory Graph links entities, files, and abstract concepts for reasoning, while Hybrid Retrieval combines FTS5 text matching with pgvector semantic similarity. Smart Extractors ingest chat logs and turn them into structured memory nodes, and Tool Connectors bring data from Claude Code, Cursor, Feishu, and GitHub into a central memory store. The page also says related decisions, projects, and people are automatically linked.

Awareness is local-first, open source, and requires no login. It offers a local mode that can be started with one command, stores data as Markdown files on the user’s machine, and works fully offline. Cloud use is optional. The page also mentions an open-source SDK on GitHub, and says cloud features can add team sync, semantic search, and multi-device access. Supported tools and environments named on the page include Claude, Cursor, ChatGPT, Claude Code, Windsurf, Trae AI, Zed Editor, VS Code Agent Mode, GitHub Copilot Extensions, PyCharm AI Assistant, OpenClaw, MCP, SDKs, and local setup flows.

The site frames Awareness as an AI cognitive memory cloud and shared memory layer for coding assistants, MCP tools, LangChain workers, and collaborative agent workspaces.