NxcoreAI released EverRoom, an open-source, local-first desktop workspace that organizes files, conversations and repositories into what it calls Context Rooms for AI agents to work from, according to its GitHub repository.
The tool pairs a replaceable AI agent runtime, compatible with OpenAI-format providers, with a memory pipeline and agent-scoped document editing, the repository says. It routes agent communication through a loopback-only local gateway and keeps data on the user's machine rather than a cloud service, according to the project.
EverRoom has drawn about 2,800 stars and 322 forks since launch, with 735 commits and connectors for Slack, Notion, Gmail and Feishu already built, the repository shows. The project describes itself as under active development, with macOS as its current target and its APIs, storage schemas and agent contracts still subject to change.
The pitch, in the project's own words, is to be "more alive than a file store, more grounded than a chat box," the repository says.
It is one more entry in a fast-growing category: tools that give an agent a durable, structured memory of a specific project instead of starting fresh each session. Whether that memory layer belongs in a desktop app, an IDE plugin or the model provider itself is still being fought out in public, one GitHub repo at a time.