AI agent payments race heats up as Visa and Tempo launch new developer tools

On Wednesday, March 18, 2026, Visa Crypto Labs released Visa CLI, an experimental command-line tool that lets artificial intelligence agents initiate card payments without embedding API keys. That same day, Stripe-backed Tempo went live on mainnet and published the Machine Payments Protocol, co-authored with Stripe, to standardize agent-to-service payments.

Cuy Sheffield, head of Visa Crypto Labs, announced on X that Visa CLI is the lab’s first experimental release. The product site says it lets AI agents make secure payments as you code and enables programmatic card transactions without managing API keys.

The project targets developers building autonomous agents that need to purchase services or data. API keys can contain sensitive credentials that agents might expose through prompts, shared repositories, or third-party tool use, creating security risks.

Tempo described its blockchain as purpose-built for payments and optimized for high-throughput stablecoin transfers, a common method for agent-led transactions. Alongside mainnet launch, the project released the open Machine Payments Protocol to give agents and services a shared way to coordinate payments programmatically.

According to the project, the spec is rail-neutral and built to extend, supporting multiple payment methods. The team lists Visa as adding support on its card network, while Stripe is enabling cards, wallets, and other options. Lightspark is supporting it over Bitcoin’s Lightning Network. In a post, Tempo predicted that agent-driven payments will soon outnumber human payments online.

Visa’s release centers on card rails and minimizing secret management for developer workflows. Tempo is promoting a multi-rail standard that covers stablecoins, cards, and other methods.

Other machine payment standards are emerging. Coinbase introduced its x402 framework in May to facilitate stablecoin payments by agents. In February, MoonPay unveiled non-custodial AI agents that would create wallets and execute on-chain transactions on a user’s behalf.

The initiatives add to industry tests of card networks, stablecoins, and the Lightning Network as rails for automated transactions tied to AI workflows.