Jack Dorsey's Block just open-sourced Berd, a desktop app for managing AI agents. The crypto media buzzed. The tech media shrugged. Both missed the point.
This is not a developer tool. This is a payment infrastructure play disguised as a desktop app. And the open-source label is a calculated distraction.
Context: The Controlled Open-Source Playbook
Block, the fintech giant behind Square and Cash App, released Berd as a desktop application for orchestrating AI agents. The announcement was sparse: no license details, no feature list, no technical specs. Just a promise of accessibility through open source.
But the crypto media's coverage, including Crypto Briefing, framed it as a positive step for AI agent management. The reality is more nuanced. Berd is a textbook case of "controlled open source" — a strategy where the source code is available but the rights to use, modify, and redistribute are carefully restricted. This is the same playbook used by Meta with Llama, Mistral with its models, and Elastic before its license change.
Why would a payment company open-source an AI tool? The answer lies in the economics of attention and transaction flow.
Core: The Real Asset Is the Payment Rail, Not the Tool
Berd's technical architecture is unremarkable. It's a desktop app — likely built on Electron or Tauri — that orchestrates LLM agents, probably supporting multiple model providers. The local-first design is a deliberate choice: it screams privacy and offline capability, but it also enables something far more valuable: local execution of agent actions, including file access and, crucially, payment initiation.
From my experience auditing DeFi protocols in 2020, I learned that the most valuable infrastructure is not the one that generates the most buzz, but the one that controls the flow of value. Uniswap's liquidity pools were the rails; the interface was just a frontend. Similarly, Berd is a frontend to Block's real asset: its payment network.
The controlled open-source strategy makes perfect sense when you realize that Block is not selling software. It is selling access to the next generation of transaction volume. Berd is a trojan horse. Developers download it for free, build agents on it, and those agents will eventually need to execute payments. Block's Square SDK and Cash App Pay are the natural endpoints. Code does not lie, but incentives often do. The incentive here is to capture the commission on every machine-initiated transaction.
This is the race for "agentic commerce" — the term I use to describe the economy where autonomous agents execute purchases, subscriptions, and microtransactions without human intervention. Stripe launched its Agent Toolkit in June 2025. Block is now countering with Berd. The battleground is not the agent management tool; it's the payment rail beneath it.
Contrarian: The Decoupling Thesis — Berd Is Not a Crypto Project, But It Will Enable One
Most analysts will compare Berd to LangSmith, Dify, or CrewAI. They will note its lack of ecosystem, its small developer community, and its limited features. They will dismiss it as a non-competitive entry.
They are correct in the short term. But they miss the decoupling: Berd's value is not in its current functionality but in its potential to bootstrap a new category of agentic commerce that is inherently crypto-native. Jack Dorsey's Bitcoin advocacy is not a coincidence. Berd is the first concrete product that aligns with his vision of an open, decentralized internet of agents — what he calls Web5.
Imagine an agent that manages your DeFi positions, negotiates with other agents for best swap rates, and executes payments over the Lightning Network. That agent needs a manager that understands both the AI orchestration and the payment settlement. Liquidity is the only truth in a vacuum of trust. Block's regulated payment infrastructure plus Bitcoin's trustless settlement layer could create a hybrid that is uniquely positioned for agentic commerce.
The contrarian angle: Berd is not a competitor to existing AI agent platforms. It is a bridge between the AI world and the crypto world. The desktop form factor is crucial — it allows the agent to interact with local wallets, sign transactions, and manage private keys without exposing them to the cloud. This is something no SaaS tool can offer.
The Risks: Open-Washing and the LLM Provider Trap
Two risks dominate. First, the controlled open-source strategy could backfire. If Block's license is too restrictive, the developer community will label it "open-washing" and reject it. Elastic and Redis lost years of community goodwill after similar moves. Block must navigate this carefully.
Second, the entire agent management layer may be subsumed by the LLM providers themselves. OpenAI, Anthropic, and Google are integrating agent orchestration directly into their APIs. Why use a separate tool when the model can manage its own tools? Berd's differentiation — local execution and payment integration — must be compelling enough to survive this platform shift.
Takeaway: Position for the Infrastructure, Not the App
For crypto investors, Berd is not a signal to buy SQ stock. It is a signal to watch the emerging agentic commerce infrastructure. The winner in this space will not be the company with the best agent management UI. It will be the company that controls the payment rails when agents start spending money.
Block is placing its bet. Stripe is placing its bet. The question is: which rail will be the default for machine-to-machine payments? The answer will determine the next cycle of value creation in both crypto and traditional finance.
Stability is a feature, not a market condition. The agent economy is coming. The question is who will cash the checks.