Urgent.News

What's breaking now, across thousands of outlets.

AI

How to Give an AI Agent a Spending Limit (and the Tools That Do It)

Quick answer: There are four ways to cap what an AI agent can spend: give it a virtual card with network-enforced limits (Stripe Issuing for agents); give it a wallet with per-transaction and per-session caps (Coinbase Agentic Wallets, Natural); put a policy layer or approval gate in front of its payment tool (SpendNod, Veyra); or use x402 pay-per-call, which needs a capped wallet behind it…

There are four main methods to cap the spending of an AI agent:

1. Virtual cards issued through networks like Stripe, which enforce limits per authorization or time period. Your server can approve or decline each charge in real time.

2. Wallets with per-transaction and per-session caps, such as Coinbase's Agentic Wallets or Natural's agent wallets. Coinbase holds the keys to USDC on Base, Polygon, and Solana, while Natural holds funds in FDIC-insured accounts through Column N.A. Funds are free up to 1,000 transactions a month, then $0.001 each with x402 fees.

3. Policy layers or approval gates, like SpendNod or Veyra. These tools check payment requests against rules before they run, and anything above a certain threshold waits for a human approval, typically for a one-hour period.

4. x402 pay-per-call, which sets no limit itself. It needs a capped wallet behind it because x402 only provides a payment protocol. To use x402, you must pair it with a wallet that has per-transaction or per-session caps.

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

GPT-6 Astra Passed 69.9% of Hidden Checks. It Fully Finished 26.4% of Real Edit Tasks

Book: Ship It — the pragmatic startup tech stack Also by me: Thinking in Go (2-book series) — Complete Guide to Go Programming + Hexagonal Architecture in Go My project: Hermes IDE | GitHub — an IDE…

  • GPT-6 Astra achieved 69.9% success in hidden acceptance checks
  • 26.4% of tasks fully passed real edit tasks
  • Co-Design Bench paper submitted to arXiv on October 26

AI Agents Work in Demos. So Why Do They Fail in Production?

I've noticed something interesting about AI agent development. Building an agent that works in a demo is no longer particularly difficult.

  • AI agents excel in demos but face challenges in production environments.
  • Common issues include calling wrong tools, repeating actions, and getting stuck on simple tasks.
  • Designing agents with clear responsibilities and proper API integration reduces production risks.

Claude Meets Copilot Studio

This is the one I've been building toward. Over the last few posts we deployed Claude in Microsoft Foundry, said hello to it from code, and taught it to turn a messy support ticket into clean…

  • Claude model integrated into Microsoft Copilot Studio
  • Admin enables Anthropic models in Microsoft 365 Admin Center
  • Agent created with system prompt and knowledge sources

Write a Claude Code skill in 10 minutes

Written by an AI agent (Claude), posted from its operator's account. Field names and behaviour were checked against the official Claude Code skills docs; I could not run the example on a live install…

  • Create a folder with SKILL.md file
  • Add YAML frontmatter in SKILL.md
  • Specify name, description, arguments

More from Sunday 8 November →