Block open-sources Berd, its AI workspace for multi-model collaboration
Desktop application integrates multiple AI models and tools with local execution
1Block's new AI workspace
Block, the technology company founded by former Twitter CEO Jack Dorsey, has open-sourced Berd, a desktop application providing a unified environment for working with various AI models and tools. Released under Apache 2.0 license, Berd offers free downloads for macOS, Windows and Linux. The project, developed by 91 contributors on GitHub, reached version 0.6.2 on August 18, marking its seventh public release.
2Locally-executed AI tool
Berd is a desktop application installed directly on users' computers rather than being browser-based. Brad Axen, Head of AI Capabilities at Block, emphasizes that Berd's value comes from working directly with local projects, files and tools. The application enables persistent projects, file attachments, AI provider configurations and automation building, all within a single interface.
3Daily work surface for AI agents
Positioned as a 'daily AI work surface', Berd allows users to initiate chats, attach files, select models and review session history from one place. According to product specifications, users can always see which project, file or model is actively shaping conversations. This emphasis on visible operational state distinguishes Berd from generic chatbot wrappers.
4Solution born from internal needs
Berd was developed to solve fragmentation issues Block employees faced when working with different AI agents. Lucinda Bell from Technology Communications at Block explains they had capable agents like Goose, Claude Code and Codex but had to navigate different interfaces. Berd provides one consistent desktop application across all models and harnesses, streamlining the workflow.
5Context preservation with persistent projects
Berd's persistent projects feature lets users return to established collections of files, instructions and agent configurations. This eliminates the need to reconstruct context for each task, allowing users to build upon previous work. Block designed Berd to be accessible beyond engineering teams, enabling users to start with conversations and add tools as needed.
6Gloopies: Visual identities for AI agents
Berd gives AI agents not just chat windows but roles, instructions, skills and recognizable visual identities. Block's animated character collections, called 'Gloopies', offer predefined personas like Berdy, Pushback and Choosey, representing different work styles. Pushback plays devil's advocate on drafts while Choosey helps narrow decisions, demonstrating specialized agent roles.
7Functional avatar design philosophy
Block emphasizes that avatar design serves functional purposes beyond adding personality. As stated in a company blog post, 'The avatar makes the agent recognizable. Its role, skills and tools make it useful.' This distinction becomes crucial when workers manage multiple specialized agents simultaneously. Visual identities serve as shorthand for otherwise invisible instruction packages.
8Professional-first design approach
Berd's product specifications carefully prevent the application from becoming a novelty. The intended personality is described as 'focused, capable, companionable,' while explicitly rejecting toy-like presentations and interfaces that hide operational state behind excessive friendliness. This reflects Berd's design for professional enterprise use.
9Architectural orchestration layer
Berd is neither a new foundation model nor a new agent runtime. Built with Tauri 2 and React 19, it communicates with Goose - Block's open-source agent framework introduced in January 2025 - via the Agent Client Protocol (ACP). Goose runs as a bundled 'sidecar' process handling the underlying agent loop while Berd manages projects, sessions and configurations.
10Goose: Model-agnostic agent framework
Goose is a model-agnostic agent framework and runtime originally focused on software development workflows. It connects large language models to files, commands and external systems through the Model Context Protocol (MCP), enabling agents to take actions beyond text generation. Compatible with both proprietary model providers and locally-run open models, Goose provides the foundation for Berd's capabilities.
11Enterprise-focused design choices
Block isn't positioning Berd as a mobile product, as its open-source collaboration platform Buzz already serves mobile needs. Berd's desktop-focused design is optimized for enterprise users managing complex AI workflows. This approach demonstrates Block's emphasis on consistent work layers being as important as incremental model capability improvements for professional use cases.
12The future of AI workspaces
Berd's open-source release represents a significant step in the evolution of AI workspaces. By providing a consistent interface across models and tools, it simplifies AI integration into daily workflows. Block's move highlights the growing importance of standardization and interoperability in the AI ecosystem. Such enterprise-focused solutions could pave the way for more efficient AI adoption in business processes.
Original article: VentureBeat — Block, Berd adlı AI çalışma alanını açık kaynak yaptı