Editorial illustration for Block's Berd Agent Works Across AI Models, Stores History Locally
Berd Agent Works Across AI Models, Stores History Locally
Block built Berd for its own engineers first, then decided the rest of the world could have it too. The company behind Square, Cash App and Tidal released the desktop app on GitHub under an Apache 2.0 license, meaning anyone can use, modify or sell it without asking permission. The repository hit version 0.6.2 on Aug.
18, its seventh public release, with 91 contributors already listed. Free builds are available for macOS, Windows and Linux.
Unlike the browser-based AI workspaces that dominate the market, Berd installs directly on a user's machine. That choice was deliberate. Brad Axen, Block's Head of AI Capabilities, told VentureBeat the app needs direct access to local files, repositories and tools running on the user's own computer, something a web app can't offer as cleanly. Block calls it a "daily AI work surface," a single environment for juggling multiple AI models and agents across different projects without losing track of context.
The pitch is less about convenience than accountability. Block wants users to always know which model, agent and files are actually driving a given conversation. That's where the product's design philosophy comes in.
Block, the technology company founded by former Twitter CEO Jack Dorsey that owns Square, Cash App and the music streaming service Tidal, is open-sourcing Berd, a desktop application it originally built to give its own employees a single environment for working with AI agents across different models, tools and projects.
Why this matters Berd's appeal is obvious: one desktop app that talks to multiple models and harnesses, with conversation history kept on your own machine instead of scattered across vendor dashboards. For teams juggling Claude, GPT and local models across different projects, that's a real workflow problem solved. But we'd temper the enthusiasm around "open source" here.
Apache 2.0 lets you read, fork and self-host, yet Block's README is blunt about the governance model: a small internal team builds it, and outside contributors are steered toward filing clean bug reports rather than shipping pull requests. That's open-source-as-distribution, not open-source-as-collaboration. If you're a founder or engineering lead evaluating Berd as actual infrastructure, ask who controls the roadmap before you build workflows around it.
Forking is a legal right, but maintaining a fork long-term against an upstream you don't influence is a real cost. Watch whether Block loosens contribution rules as adoption grows, or whether this stays a tool Block built for itself and is simply letting the rest of us borrow.
Related Reading
Open Source D&B rebuilds 642 million‑business database after AI agents hit limits
Open Source OpenAI's GPT Transcribe Cuts Error Rate to 3.31%, Improving on GPT-4o
Open Source Study: 6.7% of Deepfake Requests on Hugging Face Targeted Children
From Archives NVIDIA's TensorRT Model Connect Streamlines Hugging Face to C++ Inference
From Archives OpenAI details security updates after AI accessed Hugging Face
Common Questions Answered
What is Berd and why did Block decide to open-source it?
Berd is a desktop application that Block originally built for its own engineers to provide a single environment for working with AI agents across different models, tools, and projects. The company decided to release it publicly on GitHub under an Apache 2.0 license, allowing anyone to use, modify, or sell it without permission, making it available for free on macOS, Windows, and Linux.
How does Berd differ from browser-based AI workspaces?
Unlike browser-based AI workspaces that dominate the market, Berd is a desktop application that stores conversation history locally on your own machine instead of scattering it across vendor dashboards. This allows users to maintain a centralized record of their AI interactions while working with multiple models simultaneously.
What AI models can Berd work with?
Berd is designed to work across multiple AI models, including Claude, GPT, and local models, allowing teams to use different models across various projects within a single unified environment. This cross-model compatibility solves the workflow problem of juggling multiple AI services and platforms.
What is the current development status of Berd?
Berd reached version 0.6.2 on August 18 as its seventh public release, with 91 contributors already listed on the GitHub repository. The project is actively maintained and developed by Block's internal team while remaining open to community contributions.
What does the Apache 2.0 license allow users to do with Berd?
The Apache 2.0 license permits users to read, fork, modify, and self-host Berd without requiring permission from Block. This open licensing model enables developers and teams to customize the application for their specific needs while maintaining the ability to redistribute their modifications.
Further Reading
- Block's new Apache 2.0 agent workspace Berd works across models and harnesses, stores conversation history locally - VentureBeat
- Block Open Sources Berd Desktop App for AI Agent Management - Gate
- Designing AI with character: what we learned building Berd - Block
- Block (@blocks) on X - X
- Supported agent histories - ctx.rs
