Urgent.News

What's breaking now, across thousands of outlets.

AI

Crossmint's Toolkit Picks How to Pay. Nothing Scores Whether to Pay.

Crossmint's Toolkit Picks How to Pay. Nothing Scores Whether to Pay. Crossmint launched Agent Commerce Toolkit on October 8, 2026 — live and self-serve. One integration lets an AI agent pay with any card at any online store: Card entry renders in the developer's UI and goes straight to a PCI-compliant vault. Developers get a token, never card numbers. Eligible cards use tokenized Visa Intelligent…

Crossmint has launched its Agent Commerce Toolkit on October 8, 2026, making it live and self-serve. This toolkit allows an AI agent to pay with any card at any online store. The process is streamlined as developers get a token, not card numbers, and eligible cards use tokenized Visa Intelligent Commerce/Mastercard Agent Pay credentials with spend limits and expiry. Ineligible cards fall back to the vault.

The toolkit selects the checkout path by considering availability, reliability, and cost. It offers Shop Pay-style embedded options, Universal Commerce Protocol, or browser automation. This replaces the previous four-vendor stack developers had to assemble themselves.

However, there are two crucial aspects that the toolkit does not address. First, it does not score whether a purchase should fire, meaning whether the merchant, item, and price match what the user actually asked for. The spend limit is merely a ceiling, not a judgment. Second, it does not decide if the purchase should be escalated or blocked. It simply follows the instructions provided.

Two tests were conducted on October 12, 2026, using Crossmint-shaped instructions against a live decision gate. The first test involved $240 shoes from an unverified merchant via browser automation, which was the cheapest path by cost alone, but the toolkit escalated or blocked the transaction and logged it. The second test involved a $42 pharmacy reorder from a usual merchant with a scoped $50 card, which was approved by the instruction. Both instances were blocked by the heuristic, highlighting the gap in the current system.

The missing components are a calibrated decider that scores both the how (the execution) and the whether (whether the purchase should fire). This would provide a more discriminating approach, ensuring decisions are made with the right balance of safety and flexibility. The Crossmint toolkit is designed to help agents manage real capital, with full receipts, a decision table, and a 5-step DIY guide for setting up a gate in front of any-card checkout.

Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.

Read the original at dev.to →

More in AI

Refuse the Spike If the Assistant Nods First

A backend developer opened a blank scratch directory after a standup left one cache expiry rule unresolved. The teammate had claimed that every write should expire the whole keyspace, and the claim…

  • Assistant nods first, initiating spike
  • Spike runs for ninety minutes or ends early
  • Assistant checks wrong rule, not models

Local Receipts: a receipt organizer for freelancers, powered by local Gemma 3

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle What I Built Local Receipts is a receipt organizer for Brazilian freelancers and MEIs (micro-entrepreneurs).

  • Local Receipts organizes Brazilian freelancers' receipts without cloud data transmission
  • Powered by Gemma 3 4B open-weight model running locally via Ollama
  • Extracts vendor, tax ID, date, total, and expense category from receipts

Review an Agent PR That Fetches Model-Named Docs URLs

A maintainer found pull request 218 on a quiet Tuesday. An agent had added a docs fetch to a small coding tool. The summary promised richer answers from live documentation pages.

  • Agent added context helper feature to coding tool
  • Reviewer tested agent under controlled, offline conditions
  • Proposed strict host allowlist to mitigate security risks

More from Tuesday 13 October →