Stripe Sessions 2026 wrapped up on 30 April with the densest single-day product release the company has ever shipped. An agent wallet, an Agentic Commerce Suite co-built with Meta and Google, stablecoin payouts in 160 countries, four new chains in the settlement set, and an AI-native developer console. For any fintech developer UK, payment developer, or AI agent developer UK trying to figure out where the next two quarters of work concentrate, this is the most important release calendar entry of the year so far.
It also lines up neatly with yesterday's piece on Meta's USDC creator payouts — Stripe is the connective tissue underneath both stories, and that is not coincidence.
Link Agent Wallet — Credentials Stop Leaving the Vault
The headline product is Link agent wallet. The pitch is short: an LLM-driven agent can complete checkout on the user's behalf without ever seeing a card number, BIN, or token. The wallet sits between the agent and the merchant; the agent expresses intent ("buy this", "subscribe at this tier"), the wallet authorises within scope, and the credential stays inside Stripe.
Three engineering details make this consequential:
- Scope-narrowed authorisation. The agent receives a single-use or short-lived authority to make a specific purchase. Compromise of the agent's runtime does not leak a payment instrument.
- Multi-rail support out of the gate. Brazilian Pix, US stablecoins, with Indian UPI in the pipeline. This is not a US-only crypto experiment — it is built for the markets where local rails matter more than card networks.
- Idempotency surfaced through the API. Every agent-initiated charge has an explicit idempotency key in the wallet's protocol. This is the single biggest piece of operational hygiene for non-deterministic callers, and it is finally first-class.
For a payment developer integrating an agent runtime today, Link removes the worst design decision — "do I let the agent see a token?" — by making it impossible.
Checkout Studio and the Configuration-as-Product Bet
Checkout Studio collapses what used to be a six-team checkout build (frontend, A/B platform, fraud, observability, data, finance) into a single configuration plane. Merchants edit checkout flows without code, ship them to web/native/in-store, and watch the analytics in the same surface.
The interesting thing for engineers is the escape hatch. Custom Objects (a separate launch) lets teams add domain-specific data models — subscription tiers, marketplace splits, dynamic pricing rules — that Studio can render and that Stripe's tax/fraud layers respect. The config UI is for the 80% case; the data model is open for the rest. This is the same architectural pattern Shopify uses to keep merchants in the platform without forcing every flow through a no-code editor.
Read the full article on tomcn.uk →
About the Author
I'm Tom Wang, an AI Developer & Fintech Developer — building AI agents, crypto payment infrastructure, and cross-border payout systems with Rust, Go, and TypeScript. Based in London, UK.
Currently open to new opportunities in fintech, crypto payments, and AI agent engineering.
