Block open-sources Berd, a desktop workspace for agent crews
Block just open-sourced Berd, a canvas-based desktop environment built to coordinate specialized AI agents directly alongside your local files and repos. Users can spin up agents with distinct skills and personalities while swapping underlying models on the fly. It is a thoughtful push to move AI interactions away from disposable chat windows toward durable project building.
Berd is an open-source desktop workspace designed to bundle your conversations, local files, tools, and multi-agent crews directly into a persistent project canvas.
Its goal is simple: end the constant switching between disconnected chat windows. Instead of tossing prompts into disposable threads, Berd attaches directly to the local folders and code repositories you are already working in, anchoring your assets in one dedicated environment.
What you can do with it
Rather than relying on a single all-purpose assistant, Berd centers on deploying specialized crews. Each agent is equipped with defined capabilities and distinct traits, ready to tackle projects from multiple angles.
The system includes several built-in personas. Berdy serves as a steady guide to take work off your plate. Pushback acts as a sharp, unvarnished critic that spots flaws you might overlook. Choosey helps break indecision by clarifying tradeoffs. Copycat learns your voice to draft text that sounds like you. Tinker rolls up its sleeves to build custom skills or new agents, while Wildcard delivers unfiltered, unconventional ideas.
Demonstrated use cases range from animating logos to building interactive 404 pages with drawing tools and generating reactive slide decks. Reports also note that Berd can orchestrate workflows across goose, Claude Code, and Codex, while allowing users to doodle custom avatars for each agent.
How to get started
Getting started takes just a few steps. You link your preferred AI provider, create a project, map it to your local directories, and assemble your agent roster.
You are not locked into one vendor. You can assign demanding reasoning jobs to heavier models and swap in smaller models for mundane tasks, cutting out unnecessary compute costs. Readers can explore the tool and set it up by visiting berd.xyz.
Why it matters
If your daily work involves bouncing across prompts, documents, and terminals, Berd offers an organized alternative that preserves context locally.
We think its strength lies in moving past the generic chatbot prompt box. By packaging distinct roles, local directory bindings, and provider freedom into a persistent canvas, it turns raw model output into a coherent desktop production line.
Sources (1)
- berd.xyz(opens in a new tab)Originalberd.xyz