Mercury Spend launched on Tuesday, August 11, from San Francisco. The product lets businesses issue cards to employees and to AI agents. Intelligent budgets and self-enforcing expense policies come with them. Agent cards are virtual rather than physical, and the pitch is direct. Founders run leaner teams than ever, a growing share of their spending starts with software, and finance tooling has not kept up.

Mercury built its name as a banking alternative for startups tired of traditional business banks. Since then, it has layered on treasury and spend products. The aim is keeping customers from graduating to a dedicated corporate card provider.

Mercury Spend Arrives After Ramp on Agent Cards

One framing around Mercury Spend needs correcting first. This is not a company getting ahead of where the category is going.

Ramp launched Agent Cards in March 2026, five months earlier. Those cards issue tokenized credentials per transaction through Visa protocols, scoped to a single agent. Controls cover spend limits, merchants, approval workflows, category restrictions and full transaction visibility. Moreover, agents inherit the approval chain of the issuing user, so no purchase bypasses existing rules. Ramp followed on July 16 with AI Token Spend Management, a dashboard for tracking token and subscription costs across providers.

So the controls this story needs proving are controls a rival already shipped. Limits scoped to a task, revocation, and audit trails a controller would accept all exist elsewhere. Consequently, Mercury Spend is late on ordinary automated controls and late on the agent version too.

Mercury Spend Lands in a Market Where Agent Money Is Real

Still, scepticism about whether agentic spending exists yet deserves revisiting. Here the data cuts the other way.

Ramp reports average monthly AI token spend across its customers rising 13 times since January 2025. Cofounder and chief executive Immad Akhund says Mercury customers were already issuing virtual cards to agents by hand before this launch. That is the strongest argument in the announcement. In effect, the company is productising behaviour it observed rather than inventing a category and hoping demand follows.

Notably, this is not the first agentic move either. The company shipped a command-line interface aimed at AI agents in June. Now Mercury Spend routes every function through Command, its natural language layer.

What Mercury Spend Does Differently

The genuine Mercury Spend differentiator sits in placement rather than features. Ramp and Brex are spend management platforms that sit beside a bank. Mercury is the bank.

Mercury Spend is built into the banking dashboard. So controls, visibility and transaction data live where the money already sits. By contrast, traditional spend tools require a finance team to configure, integrate and maintain them. Because Mercury already knows cashflow and runway, budgets connect to that picture from day one.

The ordinary features underneath are solid. Teams get stipends scoped to travel, software, home office or supplies. Transactions and reimbursements route to the right budget automatically. Receipt capture runs through Gmail scanning and text reminders. None of that is novel against Ramp or Brex, but none of it requires leaving the bank either.

Mercury also released a limited-edition IO card designed with sculptor and ceramicist Simone Bodmer-Turner. It derives from an original sculpture, rendered in layered printing. That is positioning rather than product, and it tells you which buyer Mercury wants.

What to Watch on Mercury Spend

Usage disclosure is the first gap. Mercury has not said how many customers use agent cards or given a rollout timeline beyond the initial launch.

Watch the integration argument next. If being inside the banking dashboard genuinely reduces setup burden for a small finance team, that beats arriving first. Mercury serves more than 300,000 businesses and raised $200 million in May at a $5.2 billion valuation, so it has the distribution to test that claim quickly.

Watch the competitive field too. Capital One announced a deal to acquire Brex in early 2026. Meanwhile, Ramp reached a $44 billion valuation in June on roughly $1 billion in annual recurring revenue. Consequently, Mercury Spend competes against one rival being absorbed into a large bank and another scaling fast on AI-native finance tooling. Neither dynamic favours a slow follower.

The deeper question is whether agent spending stays a card problem at all. Token costs, invoices and card purchases sit in different systems by default. Whoever unifies them wins more than whoever issues the prettiest card.

For related reading, our analysis of business banking from Ramp and Mercury maps this exact rivalry. Our guide to AI in fintech tracks agentic adoption across the sector, while our piece on digital banking questions for smaller businesses covers the buyer Mercury targets. Mercury published the launch post on its blog. Fast Company covered the agent cards exclusively, and Ramp documented its own agent spend controls earlier this year.

Fintechbits covers business banking, spend management and agentic finance. Nothing here constitutes financial or investment advice. All analysis represents the editorial views of Fintechbits.