Structure complex AI work like a project, not a thread.
Darby pairs a capable AI workspace (models you choose, compare mode, multimodal input, voice, optional local models) with Darby Labs: a visual canvas where nodes form a graph, independent steps run in parallel, and outputs become memos, tables, slides, apps, images—and autonomous agents when you need depth.
Infinite canvas.
Lay out work on an infinite canvas: draggable nodes, edges, a command palette, and a node palette for deep research, chat and prompt steps, memos, lists, tables, slides, apps, images, web pages, YouTube, inputs, files, data sources, notes, frames—and Labs agents. Skill templates and quick layouts jump-start analysis, decks, reports, and more.
Autonomous agents.
Labs agents show reasoning in the thought stream and timeline, search and read the web, work with files, run sandboxed shell and Python, call HTTP APIs, produce charts and documents, generate images, structure data into tables, delegate to sub-agents on the same canvas, and pause to ask clarifying questions before heavy runs.
Deliverables, not threads.
Download outputs as a ZIP, push artifacts to GitHub, and wire automation: public execute API, inbound webhooks (e.g. Zapier), Slack notifications, MCP servers as tools, and cron schedules for recurring runs.
Planonthecanvas,shipfinishedwork
The graph you lay out tells Labs what to run and in what order. Agents return memos, decks, code, bundles, and pushes to GitHub—concrete outputs, not another endless chat. Those results feed the next iteration on the canvas.
Each node is its own run
Every node carries its own slice of the job—context, model choice, and execution—so parallel branches and reruns do not collapse into one brittle chat thread.
Recenthighlights
Product notes from building Darby Labs and the pro workspace.
Try Darby now.
You can be up and running in minutes. Organize AI work as a structured project on the canvas—not a single long thread—then export and ship real deliverables from Darby Labs.