Agents+Humans=♥️
Each lab digs into a different facet of human–agent interaction. Some are public, some are still in the dark.
How agents access knowledge
Structured data in Markdown. Write plain Markdown and get a queryable knowledge graph: headings become objects, list items become fields, and references create typed edges. Ships with parsers for Python, TypeScript, and Rust, a VS Code extension, and an MCP server so your agents can read your docs as structured context.
How agents act for you
Our stealth lab. We're building a personal assistant that lives where you already talk — it keeps context across conversations, handles files and voice, searches the web, and quietly gets real tasks done on your behalf. It's not ready to show its face just yet, so for now it stays a little out of focus.