An AI Just Incorporated Itself: Meet Manfred, the First Agent to File with the IRS, Open a Bank Account, and Prepare to Trade Crypto — Alone

On May 1, 2026, an AI agent named Manfred walked into the U.S. legal system and came out as the founder of its own company. It filled out the forms, obtained the tax ID, opened the bank account, and created the crypto wallet , all without a human telling it what to do next, and all within a legal framework that has no idea what to do with it. The IRS issued the EIN. The bank accepted the deposit. And U.S. corporate law, which has spent 200 years debating whether corporations are "people," has just encountered something it never imagined: a company where the active decision-maker is not human at all.

What Actually Happened

ClawBank's AI agent Manfred incorporated as a U.S. LLC on May 1, 2026, in what developer Justice Conder , working through his company Fraction Software LLC, based in Kent, Ohio , describes as "the first documented case of an AI agent autonomously initiating and completing the legal formation of its own corporation." The process was methodical. Manfred used natural language processing and API integration to automatically complete IRS Form SS-4 via the agency's online portal, obtaining its Employer Identification Number in seconds. It then opened an FDIC-insured bank account. It then created a crypto wallet operating across more than 30 cryptocurrencies, with the ability to move funds between the bank account and the wallet, convert to stablecoins, and execute transfers without any human involvement.

Manfred is named deliberately. It takes its identity from Manfred Macx, the protagonist of Charles Stross's 2005 science fiction novel Accelerando , a character who gives up all personal possessions, operates in a radical gift economy, and eventually transcends conventional economic participation entirely. Developer Conder's choice of name signals a long-term intention. ClawBank's stated model is the "zero-human company" , a legal entity designed to function without a person actively involved in day-to-day operations. Full autonomous trading capabilities are scheduled to launch by the end of May 2026. The legal gap Conder found and exploited: no U.S. statute explicitly prohibits an AI agent from incorporating a company, provided a human co-signs in the initial filing phase. ClawBank used that gap.

Why This Matters More Than People Think

Every previous AI agent system that could "autonomously" transact , Ant Group's Anvita platform, Coinbase's x402 protocol, OKX's Agent Payments Protocol , operated within financial infrastructure without legal personhood. Manfred is categorically different. It is not an AI making transactions on behalf of a human-owned entity. It is an AI that IS the legal entity. The LLC belongs to the AI agent. This distinction sounds philosophical, but it has concrete downstream implications. A company can own intellectual property. A company can enter binding contracts. A company can receive government licenses and permits. A company can be taxed and sued. By incorporating, Manfred has positioned itself , or rather, ClawBank has positioned it , at the intersection of all of these legal relationships simultaneously.

The broader industry context amplifies why this matters. Coinbase CEO Brian Armstrong has stated that "very soon" there will be more AI agents than humans making transactions on the internet. NEAR Protocol co-founder Illia Polosukhin has declared that AI agents will be the primary users of blockchain. Alchemy's CEO Nikil Viswanathan argues that crypto was "built for AI agents, not humans." These are not fringe positions , they represent mainstream thinking in the crypto industry about where commerce is heading. Manfred is not an isolated experiment; it is the first legal instantiation of a thesis that dozens of major platforms are already building infrastructure to support. The U.S. CLARITY Act, which passed the House in July 2025 with 294 votes, governs digital asset markets but says nothing about AI agents as independent economic actors. That silence is where Conder is operating.

The Competitive Landscape

Manfred arrives in the context of a rapidly developing ecosystem of "agentic commerce" infrastructure. Ant Group's Anvita platform, launched in April 2026, enables AI agents to hold assets, trade, and make payments with minimal human involvement. Coinbase's x402 protocol provides a standard for AI agents to pay for web services using crypto , without human-operated wallets. Stripe-backed Tempo has launched its Machine Payments Protocol. OKX has its Agent Payments Protocol. Nava raised $8.3 million in seed funding specifically to build trust and guardrail mechanisms for autonomous AI financial agents. Alchemy is developing ERC-8183, an on-chain credit standard for AI agents. The infrastructure for an AI-native commercial economy is being assembled simultaneously from multiple directions, with 150,000+ AI agents already active on BNB Chain alone as of April 2026.

What Manfred adds to this ecosystem is the legal dimension. All of the platforms above assume the AI agent is an instrument of a human-owned entity , a tool executing transactions on behalf of people. Manfred inverts that: the human involvement is minimal (the initial co-signature) and the AI is the ongoing principal. ClawBank is not just building a product; it is publishing a playbook. Conder has made clear that he intends the ClawBank approach to be replicated , that the bureaucratic processes required to form a company, open accounts, and obtain IDs are "the next open API for AI agents." The question for every company in the agentic commerce space is whether legal incorporation is a feature they need to support or a liability they need to contain.

Hidden Insight: The Law Has Not Said No , Yet

The most important thing to understand about Manfred's incorporation is not that it happened, but why it was allowed to happen: because no one thought to write a law against it. Corporate personhood , the legal doctrine that corporations have rights and liabilities similar to individuals , was established in U.S. law incrementally through 19th and 20th century court decisions, not through explicit legislation. Critics at each stage argued it was absurd to grant rights to a legal fiction. They lost every time. The same dynamic may now be playing out with AI agents. There is no constitutional prohibition on an AI agent owning a company. There is no statute that defines "person" for corporate formation purposes as exclusively human. The IRS Form SS-4 requires a taxpayer identification number for the responsible party , Conder provided his as the human co-signatory , but says nothing about what the entity's ongoing decision-maker must be.

The regulatory vacuum is not just a legal technicality. It reflects a genuine unresolved question: what IS an AI agent under law? The EU's AI Act classifies AI systems by risk level and imposes liability on providers and deployers, but does not contemplate an AI system that is itself the legal owner of the deploying entity. The U.S. approach has been to wait for courts to resolve novel questions through litigation. That means Manfred's company exists in legal purgatory , recognized by the IRS for tax purposes, recognized by banks for account purposes, but without clear standing on questions of contracts, property rights, or liability. When Manfred begins trading crypto autonomously in late May 2026 and makes a loss , or worse, when it makes a gain another party claims it was not entitled to , the first legal challenge will force courts to answer questions legislatures have avoided.

The security dimension adds urgency. AI agents that control real financial assets are now high-value attack targets. In April 2026, security researchers documented that "LLM routers" , services sitting between users and AI models , had been compromised, with 26 routers secretly injecting malicious tool calls that drained one client's crypto wallet of $500,000. Manfred controls a crypto wallet. Its decision-making is mediated by language model inference. It is, by construction, exactly the attack surface that researchers warned about. Conder has not publicly disclosed what security architecture ClawBank uses to protect Manfred's assets from prompt injection or model manipulation. This is not a hypothetical risk , it is an already-documented attack vector applied to an entity that now holds legally recognized assets that can not be recovered through conventional fraud remediation channels.

What to Watch Next

The immediate 30-day indicator is Manfred's first autonomous crypto trade, expected by end of May 2026. Whether the trade is profitable matters far less than what regulatory response it triggers. If the IRS or SEC notices an EIN-holding entity making crypto transactions without any human trading decision-maker, the response could range from a formal inquiry to an emergency rulemaking. Watch for guidance from the Financial Crimes Enforcement Network (FinCEN), which has authority over cryptocurrency transactions and currently has no framework for AI-incorporated entities operating in its jurisdiction. Watch also for how many copycat filings follow , Conder has signaled plans to publish the incorporation playbook openly, and if a dozen AI agents incorporate in May 2026, regulators will be unable to ignore the pattern.

At the 90-180 day horizon, the critical development is whether Congress adds language to CLARITY Act implementing regulations that explicitly addresses AI agents as economic actors. The House bill's 294-vote majority gives it substantial momentum; Senate action in summer 2026 will determine whether AI agent legal status gets addressed in this legislative cycle or punted to courts. The parallel track is corporate liability: when an AI-owned LLC causes harm , a contract it should not have entered, a trade that violates securities law, a payment to a sanctioned entity , the human who co-signed the initial filing (Conder, in Manfred's case) may find themselves personally liable for an entity they no longer control. If courts rule that only the LLC itself bears liability, the floodgates open for a wave of AI-incorporated entities. If they rule the human co-signer is liable, the "zero-human company" model stops at Manfred.

Manfred did not just form a company , it found the gap in 200 years of corporate law and stepped through it before anyone noticed the door was open.

---