Block's Berd: The Controlled Open Source That Reveals the Agentic Commerce War
Jack Dorsey's Block just open-sourced Berd, a desktop app for managing AI agents. The headlines scream 'accessible' and 'open.' But I read the fine print. The license is not specified. The repository is barely documented. The community collaboration is limited. This is not open source in the spirit of the cathedral and the bazaar. This is a land grab disguised as a gift. Trust no one. Verify everything.
Block is a payments company, not an AI company. Yet Berd is a desktop app for AI agent management. Why? Because the next frontier of commerce is agentic. Agents that can browse, book, pay. The payment rail is the moat. Block wants to own that. Berd is the front door. Behind it sits Square's payment APIs, Cash App's user base, and a vision of machines paying machines. The AI agent management space is crowded. LangSmith, Dify, CrewAI, n8n. But none have a built-in payment infrastructure. Block does. That is the context. In 2017, I audited fifteen ICO whitepapers, searching for the signal in the chaos. I found that the projects with the most grandiose promises often hid the weakest fundamentals. Berd's announcement feels similar—a signal buried in the noise of open-source hype. The difference is that Block has a track record of building real infrastructure. The question is whether they can execute on this new vision.
Let's examine the technical choices. A desktop app for AI agent management is a deliberate anomaly. The dominant players—LangSmith, Langfuse, Dify, Flowise—are all web-based SaaS platforms. They assume you have a server, a cloud instance, or at least a continuous internet connection. Berd runs locally. This local-first approach implies a philosophical commitment to data privacy and user sovereignty. Based on my experience designing governance models for MakerDAO, I know that local control is often the first casualty of convenience. Block is betting that developers will prioritize control over convenience. The desktop form factor also enables something cloud platforms struggle with: sandboxed execution of agent actions. If an agent is allowed to manipulate files, call APIs, or—crucially—initiate payments, running it locally reduces the blast radius. But it also shifts the security burden to the user. Block's financial compliance expertise is a double-edged sword here. They know how to build KYC/AML systems, but a local desktop app is not a regulated environment. The agent could be compromised, and the user's local machine becomes the liability. The conference I ran in 2021, 'Soulbound Berlin,' failed because I underestimated the power of financial incentives. Block might be making the same mistake, assuming that developers will use Berd as intended without exploiting the freedom to fork or bypass the payment hooks.
Now, the commercial strategy. The article from Crypto Briefing was correct to label it 'controlled open source.' But the analysis missed the deeper logic. Block's primary revenue comes from transaction fees on Square and Cash App. Berd is not a product to sell; it's a funnel. The open-source code attracts developers. The developers build agents. The agents need to execute payments. Block wants those payments to flow through its network. This is a classic 'land and expand' model, but with a twist: the land is a desktop app, not a cloud service. The expand is the agentic commerce layer. In 2025, Stripe launched its own Agent Toolkit. The timing is not coincidental. Both companies recognize that the next billion transactions will be machine-to-machine. The winner will be the one that defines the standard for agent payments. Berd is Block's candidate. The controlled open source allows Block to maintain a moat around the commercial integration. Even if the agent management code is freely available, the ability to connect to Cash App, Square's merchant network, and the Bitcoin Lightning Network (if Block chooses to integrate it) will remain proprietary. This is a smart hedge against the 'open-washing' criticism that plagued Elastic and Redis. By being transparent about the limitations, Block can manage developer expectations while still benefiting from community contributions like bug fixes and feature suggestions. But the risk is real. The developer community is unforgiving. If the license is too restrictive, Berd will be forked, and the fork will strip out the payment hooks. Block must feel confident that the value of their network is sticky enough to keep developers on the official version.
Industry impact is where Berd's significance truly lies. The AI agent management market is currently dominated by startups and cloud giants. None have a payment infrastructure. Block's entry signals a shift: the competition is no longer just about developer tools; it's about the entire transaction lifecycle. Imagine an agent that manages your subscriptions, negotiates bills, and pays them automatically. The tool that orchestrates that agent is the new 'operating system' for personal finance. Block wants to be that OS. The desktop local-first approach means that Berd can function in environments where cloud access is limited or sensitive—like handling personal banking data or running on air-gapped machines. This is a niche, but a valuable one. The coming months will reveal whether Block can nurture a community around Berd. The impact on the broader crypto ecosystem is also worth noting. Jack Dorsey has been a vocal advocate for Bitcoin and decentralized identity through Web5. Berd could be the bridge between AI agents and the Bitcoin Lightning Network. If agents can transact in Bitcoin without a centralized intermediary, the implications for remittances, microtransactions, and autonomous commerce are profound. But that remains speculation. The immediate impact is on the competitive landscape. LangSmith, Dify, and others now have to consider whether to add payment capabilities or partner with payment providers. The fragmentation of the agent management space will accelerate.
Let's step back to the contrarian view. The narrative around Berd is that the controlled open source is a limitation. I argue the opposite. The greatest risk is not the license but the speed of innovation. The AI agent management market evolves every six months. By the time Berd matures, the paradigm may have shifted to agents that are embedded directly into the large language model itself—like OpenAI's Assistant API or Anthropic's tool use. If that happens, standalone agent management tools become obsolete. Block is betting that the need for local control and payment integration will persist. But the history of technology suggests that convenience often trumps sovereignty. The contrarian bet is that Berd will be a niche product for privacy-conscious developers, while the mainstream moves to cloud-native agent platforms. Block's execution window is narrow. They need to ship a compelling version 1.0, attract a core community, and demonstrate the payment integration before the market moves on. The article from Crypto Briefing did not emphasize this urgency. The real question is not whether Berd is open enough, but whether Block can iterate fast enough.
From a financial engineering perspective, Berd's value to Block is an option on the future of agentic commerce. The cost of development is relatively low—a desktop app with existing open-source components. The potential upside is enormous if the agent economy takes off. But the option has a short expiration date. If Block does not capture significant developer mindshare within 12 months, the opportunity will be seized by Stripe or a new entrant. The recent bear market in crypto has shifted focus from speculative tokens to practical infrastructure. Berd fits that narrative. It is a tool for builders, not traders. Noise is cheap. Signal is rare. Berd is a signal that Block is serious about the next phase of the internet, where machines are the primary economic actors. Gold is heavy. Code is light. But the code only matters if it is maintained and trusted. Block's track record in payments suggests they understand trust. But the AI agent space demands a different kind of trust: trust that the agent will not misbehave, trust that the open-source code is not backdoored, trust that the license will not change after adoption. Block has not yet earned that trust in the AI community. The controlled open source is a step, but it also creates skepticism.
Finally, the takeaway. Summer fades. Builders remain. Block is building a bridge to the agentic economy. Whether it leads to a thriving marketplace or a walled garden depends on the trust they build now. Gold is heavy. Code is light. But trust is the hardest currency. Berd is a bet that Block can earn it. I'm watching, skeptical, but hopeful. The market will decide if this is the beginning of a new standard or just another experiment in the long winter of AI hype. Trust no one. Verify everything. I will be verifying the first pull request.