Most believe open source is a gift. A benevolent act of sharing. It is not. It is a strategic weapon, deployed at the precise moment when the battlefield shifts.
Block, Jack Dorsey's financial infrastructure giant, just open-sourced Berd—a desktop application for managing AI agents. The crypto media reported it as a measured step: "controlled open source," limited community collaboration. They missed the point. This is not about code. It is about control.
Context: The Quiet Invasion of the Desktop
Berd is a desktop app. In an industry dominated by web-based SaaS platforms like LangSmith and Dify, Block chose local-first. Why? Because desktop means three things: offline resilience, local data sovereignty, and the ability to execute actions on the machine itself. For an AI agent, that means it can manipulate files, trigger local scripts, and—most critically—initiate payments.
Block is not an AI company. It is a payment company. Its core revenue comes from Square merchant fees and Cash App transaction volumes. The company's strategic DNA is to build infrastructure, attract developers, and monetize the flow of value. Berd is the latest iteration of this playbook.
Consider the market context. In 2025, the AI agent economy is still embryonic. But the underlying race is not about building better agents—it is about owning the rails through which those agents transact. Stripe launched an Agent Toolkit in June 2025. Now Block has countered with Berd. The two payment giants are positioning themselves as the gatekeepers of agent commerce.
Core: The Architecture of the Trap
Berd is a combination of existing technologies: desktop GUI frameworks, agent orchestration logic, and a local-first data layer. But its true innovation is not technical—it is commercial. The app is designed to be a sandbox for agent behavior, and that sandbox sits on a device that can access the user's financial accounts.
Based on my audit of payment infrastructure codebases, I can tell you that the most dangerous feature is the one not yet announced. Berd almost certainly has hooks for Square's payment SDK and Cash App Pay. The open-source version allows developers to build and test agents locally. But the commercial version—the one that connects to Block's payment network—will be a proprietary layer. This is the classic "controlled open source" strategy: give away the razor, sell the blades.
Scarcity is a narrative; utility is the anchor. The desktop form factor is utility. It allows agents to execute real-world tasks: booking a flight, ordering groceries, paying a subscription. Each of these actions requires a payment gateway. Block owns the gateway.
Let's examine the competitive landscape. LangSmith has the largest ecosystem. Dify is open-source but cloud-native. OpenAI's Agents SDK is powerful but locked to their API. None of them have native payment integration. Berd's unique selling proposition is that it can become the default operating system for agents that need to spend money.
Consensus is often just coordinated delusion. The market consensus is that agent management tools are a developer choice. I believe they are a strategic asset. The winner will not be the tool with the most features, but the one that controls the transaction flow.
Contrarian: The Controlled Open Source Paradox
Critics label Berd's license as "controlled open source"—a pejorative term implying restricted freedom. But this ignores the reality of public company strategy. Elastic, Redis, and HashiCorp faced similar backlash when they moved to restrictive licenses. Yet each survived because the value of their ecosystem outweighed the community's anger.
Berd's license is not a failure of generosity. It is a necessary firewall. Block cannot afford to let competitors build rival payment layers on top of its open-source code. The license ensures that the only path to monetization for agents built on Berd is through Block's infrastructure.
Hype decays; adoption endures. The real risk is not community backlash. It is the possibility that the entire agent management layer becomes irrelevant. If OpenAI or Anthropic embed agent orchestration directly into their model APIs, independent tools like Berd face obsolescence. Block's hedge is its payment network. Even if the management layer is commoditized, the transaction flow remains.
Takeaway: The Battle for the Agent's Wallet
Berd is not a product. It is a positioning statement. Block is saying: "We will be the bank for your AI agents." The success of this strategy depends on whether developers trust a payment company with their agent's autonomy.
In my years modeling token emission schedules, I learned one thing: the most valuable asset in any network is the ability to capture value at the point of transaction. Block is placing a bet. The market will decide if it's a bridge or a bottleneck.
Watch the developer adoption. Watch the license updates. And watch for the first agent that pays for a coffee using Cash App.