we’re building ars umbris around the idea that a knowledge base should work like a codebase
your notes, types, agent skills, tools and UI components live together in local repos. a type engine checks fields, sections and references, so agents get feedback when something is missing or broken. coding agents ( like claude code or codex ) connect through mcp and work on the same files
you can build your own tools and views or reshape the app with your agent. we built it with the same primitives and SDKs you can use. repos can depend on each other to share types, knowledge and capabilities
its all local markdown, yaml and code files on your machine
macos only for now and you’ll need to build from source. would love to hear what you think
arscontexta•51m ago
your notes, types, agent skills, tools and UI components live together in local repos. a type engine checks fields, sections and references, so agents get feedback when something is missing or broken. coding agents ( like claude code or codex ) connect through mcp and work on the same files
you can build your own tools and views or reshape the app with your agent. we built it with the same primitives and SDKs you can use. repos can depend on each other to share types, knowledge and capabilities
its all local markdown, yaml and code files on your machine
macos only for now and you’ll need to build from source. would love to hear what you think