Block Apache Agent Workspace Berd Brings Local History

A single choice—keeping every chat local—changes how safe an AI workspace feels. Block’s new agent workspace Berd is built to store conversation history on the user’s machine, according to coverage of the launch on Wednesday, August 19, 2026.

Block, the Jack Dorsey-founded company behind Square and Cash App, has released an open-source agent workspace called Berd. Coverage on Venturebeat describes Berd as Apache 2.0–licensed, model-agnostic, and focused on local-first conversation history storage. Now, the exact feature list we’re working from comes from published coverage, not official release notes we could access. But the core points—licensing, local storage, and multi-model support—are consistent across reporting.

Overview: What Block Apache means for your AI workspace today

Here’s the part that affects daily usage: the workspace is engineered to store conversation history locally. That matters when you’re running agents for coding, research, or customer support drafts, because it shifts data control away from a remote service boundary. Worth noting: local storage also means you’ll want to manage backups, disk limits, and device encryption like you would for any personal workspace.

Key Details: The three metrics that define Berd’s approach

Berd’s value proposition boils down to three measurable design goals: licensing openness, interoperability across model providers, and local-first history handling. According to coverage, these are core engineering constraints, not optional add-ons. Berd’s design targets (as described in coverage)

• Apache 2.0 license for the workspace codebase

• Local storage of conversation history on the user’s machine

• Works across models and harnesses, supporting different ways to run agent logic

Quick breakdown: what you get and why it matters

Context: Why “local-first” is becoming the default expectation

When agent tools store everything remotely, teams end up paying for convenience with compliance work. For wider coverage, see The Verge.

Local-first history doesn’t remove governance needs, but it can reduce the gap between “we used an AI tool” and “we can explain where the data went.” That’s especially relevant in India where many teams handle client documents and internal notes under stricter internal controls.

This direction also aligns with how the agent ecosystem is maturing. Major AI platforms increasingly support modular execution across different model backends, and the workspace layer becomes the glue: prompts, tools, memory policies, and run logs.

In that world, Berd’s workspace framing is a logical move—keep orchestration flexible while keeping user conversation history under the user’s control.

What’s Next: Where teams should look before adopting Berd </h2

Related Articles

• Fal Launches Fal Agent in 2026 to Orchestrate Image, Video
• Claude Haiku 5 Rumors: What's Actually Confirmed vs. Pure Speculation
• iOS 26.6.1 Update: What Apple Added for iPhone Today

FAQs

How does block apache store user conversation history?

The system stores conversation history locally on the user device rather than sending the data to external cloud servers. Jack Dorsey and the engineers at Block designed the workspace this way to ensure maximum privacy and data security for developers.

Which artificial intelligence models and harnesses can developers use with block apache?

Developers can integrate a wide variety of foundation models and development harnesses into the platform due to the flexible Apache 2.0 open-source license. Bob and his engineering team built the workspace to remain completely model-agnostic, allowing seamless switching between different AI providers.