Hook
Block, the company behind Square and Cash App, just dropped Berd — a desktop app for managing AI agents. And they open-sourced it. But if you think this is just another developer tool, you're missing the real play. The code is out, but the license is "controlled open source." The crypto media is buzzing, the tech press is silent. That tells me something: this isn't about developers. It's about turning every AI agent into a payment terminal. Code doesn't care about your feelings. It cares about incentives. And Block's incentives are clear: own the transaction layer for the agent economy.
Context
Block (NYSE: SQ) is a payments and financial services company built on the Square merchant ecosystem and Cash App consumer wallet. Jack Dorsey has been vocal about Bitcoin, open protocols, and decentralized identity (Web5). Berd is their first major AI product aimed at developers — a desktop application that lets you orchestrate, monitor, and manage AI agents. It's hosted locally, not in the cloud. That's a rare design choice in 2025, where every agent framework is SaaS-first. According to the sparse announcement, Berd is open source but with restrictions that limit community collaboration. The article I read called it a "controlled open source strategy." I call it a bait-and-switch for the agent economy. Yield is the bait, rug is the hook.
Core
Let me cut through the marketing. Berd is not a breakthrough in AI. It's a distribution channel for Block's payment infrastructure. The desktop form factor is strategic: it allows agents to run locally, access local files, execute local tools, and — crucially — initiate payments. Based on my experience auditing DeFi protocols, I've seen how a single entry point can control the flow of value. Berd is that entry point for agents. The open source code can be audited, but the controlled license means Block can gatekeep the connection to their payment APIs. That's the moat.
From a technical standpoint, Berd is likely built on Electron or Tauri, supporting multiple model providers (OpenAI, Anthropic, maybe local via Ollama). The real value is in the agent session management, tool calling visualization, and local data persistence. But the hidden feature is payment integration. Square SDK, Cash App Pay, even Bitcoin Lightning Network — if an agent can trigger a payment, Berd becomes the switchboard for agentic commerce. Every time an agent pays for a subscription, books a flight, or buys a coffee, Block takes a cut. That's a recurring revenue stream that no other agent management tool (LangSmith, Dify, CrewAI) can offer.
I've seen this play before. In 2020, Uniswap V2 liquidity mining looked like a yield opportunity, but the real money was in the fees. Same here: Berd looks like a developer tool, but the real money is in the transaction volume. The controlled open source is a classic strategy: let the community debug and extend the free version, then sell the enterprise tier that includes payment hooks. Panic sells, liquidity buys. Block is buying liquidity in the agent ecosystem by giving away the tool.
Contrarian
Most commentators will focus on the "controlled open source" as a negative — it's not true open source, it limits community participation. They're wrong. The real issue isn't the license. It's the security and trust model for agent payments. If Berd allows agents to execute arbitrary tools without proper sandboxing, we're looking at a disaster waiting to happen. I've seen smart contract exploits that drain millions in seconds. An agent with a payment API could be weaponized. Block's compliance team (AML, KYC, fraud detection) is their biggest asset, but also their biggest liability. If a rogue agent uses Berd to make unauthorized payments, who's liable? The developer? The user? Block?
Furthermore, the crypto community's excitement is misplaced. Jack Dorsey's Bitcoin focus means Berd might support Lightning Network payments, but that doesn't make it a crypto project. It's a fiat-on-ramp in disguise. The real contrarian bet is that Berd will fail to gain developer traction because the AI agent space moves too fast. LangChain and OpenAI are already building agent orchestration into their core offerings. Berd could be obsolete before it reaches critical mass. Survival is the only alpha.
Takeaway
Block is betting that the agent economy will need a payment rail, and they want to be the default. Berd is the Trojan horse. Whether it succeeds depends on two things: first, can they ship the payment integration before Stripe or PayPal does? Second, can they earn the trust of developers who are skeptical of a payments company building AI tools? The clock is ticking. In six months, we'll either see Berd integrated into Square's merchant dashboard, or we'll see it abandoned. Either way, the signal is clear: the battle for agent commerce has begun. Whoever controls the agent's wallet, controls the future of commerce.