llms.blog
AI

Block Open-Sources Berd: Apache 2.0 Desktop Workspace for Multi-Model AI Agents

Block has open-sourced Berd, an Apache 2.0-licensed desktop application designed to serve as a unified workspace for managing AI agents across different foundation models, toolsets, and execution harnesses. Originally built for internal use across Square, Cash App, and Tidal, the desktop client reached version 0.6.2 on August 18, 2026, with builds available for macOS, Windows, and Linux. The release addresses growing operational fragmentation as developers juggle specialized agent environments

axc
axc
2 min
Block Open-Sources Berd: Apache 2.0 Desktop Workspace for Multi-Model AI Agents

Block has open-sourced Berd, an Apache 2.0-licensed desktop application designed to serve as a unified workspace for managing AI agents across different foundation models, toolsets, and execution harnesses. Originally built for internal use across Square, Cash App, and Tidal, the desktop client reached version 0.6.2 on August 18, 2026, with builds available for macOS, Windows, and Linux.

The release addresses growing operational fragmentation as developers juggle specialized agent environments like Anthropic's Claude Code, OpenAI's Codex, and Block's own Goose. Rather than offering another foundation model or standalone agent runtime, Berd acts as a desktop orchestration layer that manages persistent project context, tool configurations, and agent states on top of existing harnesses.

Berd Desktop Architecture and Sidecar Communication

Sidecar Runtime and Open Protocols

Berd is built with Tauri 2 and React 19, operating entirely as a local desktop environment rather than a web application. Its architecture relies on two key open standards:

  • Agent Client Protocol (ACP): Berd communicates with its execution engine through ACP. Goose, the model-agnostic runtime that Block contributed to the Agentic AI Foundation (AAIF), runs as a bundled sidecar process to handle the core agent execution loop.
  • Model Context Protocol (MCP): Tool connectivity, filesystem access, and external data sources hook into the runtime using MCP standards, allowing users to attach tools across local and remote environments.
  • Deterministic Version Pinning: The desktop build system locks specific Goose backend binaries via lockfiles to prevent silent runtime drift, while allowing developers to point to custom local builds for testing.

Local-First Storage and Explicit State

Unlike cloud-hosted AI dashboards, Berd prioritizes data residency and operational transparency:

  • Local Data Retention: Conversation logs, project workspaces, and session histories are stored locally on the user machine by default.
  • OS Keychain Integration: API keys and provider credentials are kept in the operating system credential manager rather than synchronized to remote servers.
  • Visible Operational State: The user interface explicitly presents active project boundaries, selected models, attached MCP tools, and failure states rather than abstracting runtime errors behind conversational wrappers.
  • Visual Agent Personas: To prevent confusion when managing multiple concurrent agents with varying system prompts and tool permissions, Berd introduces visual identities (termed "Gloopies" internally) with distinct behavioral roles such as adversarial review, decision narrowing, and stylistic matching.

Roadmap and Enterprise Strategy

While Berd targets individual, single-player desktop workflows, Block maintains a complementary open-source collaboration platform called Buzz. Built on the decentralized Nostr protocol, Buzz provides shared rooms where human team members and cryptographically identified AI agents collaborate on shared repositories.

Block confirmed that it plans to migrate key single-player workflows from Berd directly into Buzz to support seamless handoffs between private local execution and multi-user environments. In the interim, Berd remains freely available on GitHub for organizations seeking an open-source, multi-harness desktop interface.

Sources

AILLMAgentsOpen SourceTooling
ShareX (Twitter)LinkedIn

Written by

axc

More to read

  • Rank Collapse in Deep Transformers: Why Pure Attention Degenerates Doubly Exponentially and How Skip Connections Preserve Capacity

    When the Transformer architecture was introduced in 2017 with the seminal paper "Attention Is All You Need", the central thesis was that recurrence and convolution could be completely discarded in favor of stacked self-attention mechanisms. However, theoretical analysis has shown that the title's premise is mathematically incomplete. Stacking pure self-attention layers in isolation does not produce an expressive deep model: it triggers a catastrophic failure mode known as rank collapse. In a fo

    1 min

  • Item Response Theory Audit of 192 LLMs Exposes Safety Benchmark Redundancies, Over-Refusal Distortions, and Sandbagging

    A psychometric evaluation of 192 frontier and open-weight language models across eight major safety benchmarks has revealed structural flaws in current safety testing methodologies. The research, conducted by Joshua Fonseca Rivera, Neil Shah, David Demitri Africa, and Konstantinos Voudouris with support from the UK AI Security Institute and the UK Department for Science, Innovation, and Technology (DSIT), applies Item Response Theory (IRT) to analyze 5,255 evaluation items. The findings demonst

    1 min

  • Binary Quantization and Matryoshka Embeddings in Production: Two-Stage Rescoring, Hamming Distance Acceleration, and Vector Storage Economics

    Production vector search workloads face a severe memory scaling bottleneck. In retrieval-augmented generation (RAG) and semantic search architectures, indexing tens of millions of high-dimensional embeddings in raw 32-bit floating-point (FP32) format exhausts memory capacity and drives up infrastructure costs. Indexing 100 million 1536-dimensional vectors requires 614.4 GB of RAM purely for vector payloads, before accounting for graph edges and memory fragmentation in Approximate Nearest Neighbo

    1 min