OpenViking
Filesystem-style context store for agent memory, knowledge, and skills
What it is
OpenViking organizes what an AI agent knows into a virtual file tree addressed by viking:// URIs, keeping documents, extracted memories, and skill definitions side by side. Agents browse it with commands like ls, read, and grep, and each directory carries generated summaries at several levels of detail so agents can check relevance before loading full content. Semantic search can be scoped to a subtree, and committed sessions are turned into editable Markdown memories. Developers building agents use it to give them persistent, inspectable context, with a web studio for browsing and editing.