Designing AI with Character: Lessons from Building Berd

At Block, the team often builds the tools they need to get their work done. Berd is a desktop application that helps teams work with AI agents across different projects, skills, tools, and models. Today, the company is making Berd open source so others can inspect, use, and adapt the application and the design lessons behind it.

Benefits

Building Berd helped the team solve a critical design question: How do you make something as abstract as an AI agent easier to understand and shape? The answer was to give agents character. This means using roles, instructions, skills, and tools alongside distinctive visual identities.

This approach builds on Block’s history of bringing inspired design to categories not traditionally known for it. Square transformed payment hardware from something sellers hid behind the counter into something they could display proudly. Cash App brought personality and cultural relevance to personal finance. With Berd, the team brought the same instinct to AI. The goal was not to disguise the technology, but to make its capabilities more visible, approachable, and personal.

Berd grew from a practical problem inside Block. The team had access to increasingly capable agents through tools such as goose, Claude Code, and Codex, but working with them meant navigating different interfaces, configuration systems, and ways of managing context. The technology was powerful, but the experience around it was fragmented, and often assumed a level of technical fluency that many people did not have. Berd gave teams one desktop application for working across models and harnesses. It brought conversations, files, folders, instructions, agents, and skills together around persistent projects. Instead of rebuilding context for every task, people could shape agents around recurring ways of working and return to that work later.

This was designed to make agentic work more approachable beyond engineering. People could begin with a conversation, understand which context and tools were active, and add more structure as the work required it.

Many AI interfaces begin with an empty prompt box. The model may be capable, but the product gives people little sense of how the agent is configured, which context and tools are available to it, and how it differs from another agent. In Berd, character is both visual and functional. Different agents can look distinct because they are distinct. They can help you expand your thinking or narrow down on a solution, write in your unique style, or build your own tools using Berd. Just by chatting with Berd, users can easily create custom agents to do any task they want, even helping with everyday tasks like planning travel or shopping.

To help people build and share their agents, the team designed distinctive collections of animated characters, including their flagship Gloopies. These designs give abstract configurations a recognizable visual identity that people can understand and have fun with. More than just giving a playful twist to the agent workflow, this approach makes creating and customizing personal agents more accessible for anyone. The avatars make the agent recognizable. Its role, skills, and tools make it useful.

Use Cases

Berd connects to goose through the Agent Client Protocol. This separation lets Berd focus on the desktop experience, including projects, context, sessions, agents, and configuration, while goose handles the underlying agent loop. Berd is a desktop application built around goose, which remains the open agent framework and runtime.

Building Berd also helped the team understand where a private desktop experience stops. Work with an agent often begins alone. You research, experiment, gather context, and shape an idea before it is ready for other people. Berd gave them a place to explore that individual experience. But work rarely stays private. Eventually, you may need to bring in a teammate, add another agent, share an artifact, explain a decision, or keep a record of how something was made. Those are collaborative problems, and they are the focus of their recently released Buzz. Buzz is their open source workspace where people and agents work together in shared rooms, on a relay the community can control. People and agents have their own identities, participate in the same conversations, and contribute to the same searchable record. What they learned from Berd will inform their continued work on Buzz. Berd showed them the importance of private space, durable context, recognizable agent identities, reusable skills, visible configuration, and clearer visibility into an agent’s configured context, tools, and capabilities. Buzz gives those ideas somewhere to go when individual work becomes collaborative. Start alone, then go multiplayer.

Pricing

No pricing details are available in the provided article.

Vibes

No reviews, testimonials, or public reception details are available in the provided article.

Additional Information

Berd grew from what the team learned building and using goose, the open source AI agent framework Block introduced in January 2025. goose connects language models to tools and real-world actions through an open, modular architecture. In December 2025, Block joined Anthropic, OpenAI, and others to establish the Model Context Protocol (MCP) under the Linux Foundation. They later contributed goose to the foundation, giving it a vendor-neutral home alongside the Model Context Protocol and other standards. Berd and goose now serve different but connected purposes. goose remains the open agent framework and runtime. Berd is a desktop application built around it. As always, the team wants to hear from users. They encourage people to download Berd, try it with their own agents and workflows, and share their feedback. They want to see what users make, hear what works for them, and understand what they think agent experiences should become.