Block open-sourced a desktop app for AI agent management. The market yawned. It shouldn't have.
Jack Dorsey's fintech behemoth dropped Berd into the open-source pool with a controlled license. No hype. No multi-million dollar marketing campaign. Just a binary โ source available, but community collaboration is restricted. The immediate reaction from developer circles was muted. A payment company building an AI tool? What do they know?
They know something the market is missing. Berd is not a tool. It is a toll booth.
Context: The Agent Management Landscape
By 2026, the AI agent management race is a crowded field. LangSmith, Dify, Flowise, CrewAI โ all cloud-first or SDK-based platforms. They compete on features: orchestration, observability, multi-agent coordination. None of them control the transaction layer. None of them own the payment rails.
Block does. Through Square and Cash App, Block processes hundreds of billions in annual payment volume. Jack Dorsey's obsession with Bitcoin adds another dimension: a decentralized, permissionless settlement layer. Now they release a desktop app that manages AI agents โ locally, offline-first, with a license that says "we control the commercial use."
This is not a developer tool. This is a strategic asset.
Core: The Controlled Open-Source Strategy
Let's strip the narrative. The term "controlled open source" is a polite way of saying "we want the community's innovation without giving away our moat." Block's license for Berd is not OSI-approved open source. It's source-available with restrictions. The industry calls this open-washing. I call it rational.
Why? Because Berd's real value is not in the code. It's in the integration.
Berd is a desktop application. That choice is deliberate. Desktop means local data persistence. Local data means the agent's conversation history, tool call logs, and โ critically โ any financial credentials stay on the user's machine. This is a privacy-first architecture that cloud platforms cannot offer. But more importantly, desktop apps are the ideal sandbox for agents to execute actions on behalf of users. File system access. Browser control. And yes, payment initiation.
Block's payment infrastructure (Square SDK, Cash App Pay) is the natural next integration. Imagine an agent that books a flight, selects a seat, and pays for it โ all without the user logging into a payment portal. The agent uses Berd to manage the workflow, and Block's APIs to execute the transaction. The transaction fee flows to Block.
This is the agentic commerce thesis in practice.
The controlled license ensures that no competitor can fork Berd and offer their own integrated payment service. Community developers can fix bugs and add features, but they cannot monetize the agent's ability to transact without Block taking a cut. This is not generosity. It is a toll booth disguised as open source.
Contrarian: The Decoupling Thesis
Conventional wisdom says Berd will compete with LangSmith and Dify on features. It will lose. The developer ecosystem around LangChain is massive. Berd's desktop-first approach feels like a step backward in a cloud-native world.
That is the wrong frame.
Berd is not competing in the developer tools market. It is competing in the agent-to-payment infrastructure market. That market does not yet exist at scale. But the trajectory is clear. Autonomous agents will execute transactions on behalf of humans. By 2028, I project that AI-driven agents will account for 15% of all digital trading volume โ a claim I made in my 2026 liquidity synthesis report. If those agents need a payment gateway, the platform that controls that gateway captures a recurring fee on every machine-to-machine transaction.
Block's bet is that Berd becomes the default agent management layer for users who want their agents to spend money. The local-first design means agents can operate securely without constant cloud oversight. The controlled license means Block retains the exclusive right to connect those agents to payment rails.
The decoupling is this: Berd's success will not be measured by GitHub stars. It will be measured by transaction volume.
If Berd integrates Bitcoin Lightning Network โ and Jack Dorsey's history suggests it will โ then we are looking at a bridge between the agent economy and the crypto economy. Agents could autonomously open Lightning channels, route payments, and settle in Bitcoin. That is a paradigm shift. Traditional fintech is not built for machine-to-machine micropayments. Bitcoin is. Block is positioning itself at the intersection of two exponential curves: AI agents and decentralized finance.
Takeaway: Positioning for the Next Cycle
Bear markets are for building. Berd is a build. It is a small, unassuming desktop app that will either fade into obscurity or become the operating system for agentic commerce. The market is not pricing this possibility.
Liquidity vanishes. Code remains.
Regulation doesn't stop innovation. It taxes it.
Bears don't build. They just short.
Block's move is a long option on the agent economy. The controlled open-source strategy is a hedge against the risk of giving away too much. The desktop-first approach is a bet on privacy and local control. The payment integration is the ace.
Will agents become the next major liquidity source? If they do, Berd will be the toll booth. And you will remember this article.
--- Based on my experience auditing DeFi liquidity during the 2020 crash and modeling CBDC impacts on private sector liquidity in 2022, I see the same pattern: infrastructure built in the bear market captures the bull market flows. Berd is that infrastructure for a world where machines pay machines.
