Newshot

Claude Linked Billing Test Goes Live, Enabling "Pay-as-You-Go" Settlement for AI Usage in Third-Party Apps

AI summary

Anthropic updated its Claude platform documentation on November 17, introducing an OAuth beta feature called "Linked Billing." Users will be able to pay for the actual AI calls behind a third-party AI app directly through their own Claude account. Developers can link their app to a customer's Claude Platform organization, letting the customer pay directly for the Claude API calls made in the app, instead of the developer bearing the model usage costs. Users must hold an Admin or Billing role in their Claude organization to authorize, can view app costs and set monthly spending limits on the "Linked Platforms" page in Claude Console, or unlink at any time; apps in the beta phase require a monthly limit of no more than $50. Anthropic stated that requests actually run in the developer's own organization, so the developer's rate limits, model access, service tiers, data retention policies, and encryption keys apply, while the customer's zero data retention, business associate agreements, or customer-managed encryption keys do not automatically apply to these requests.

Translation(model translation, not the source's own text)

Claude Linked Billing Test Launches, Third-Party App AI Usage Can Be "Billed On Demand"

IT之家, November 18 — Anthropic yesterday (November 17) updated its Claude platform documentation, announcing the introduction of an OAuth beta feature called "Linked Billing." In the future, when users use a third-party AI application, they will be able to pay the actual underlying AI invocation costs directly with their own Claude account.

In the past, when enterprises or developers called Claude from third-party tools such as coding assistants or workflow automation, they often had to share API keys. Once a key leaked, others could use it to make calls and incur charges.

With "Linked Billing," developers no longer bear model usage costs uniformly. Instead, developers can link their own application to a customer's Claude Platform organization, letting the customer pay directly for the Claude API calls generated in the application.

In terms of control and security, users must hold the Admin or Billing role in their Claude organization to complete authorization. Users can view application costs on the "Linked Platforms" page in the Claude Console, set a monthly spending cap, or unlink at any time; for applications in the testing phase, users must set a monthly limit of no more than 50 US dollars (IT之家 note: roughly 334 yuan at current exchange rates).

For small and medium-sized AI application teams, this feature is expected to change the traditional business model of "the platform pays upfront and then resells to customers." Customers only need to click "Link your Claude organization" within the application, log in to Claude and complete authorization to bind the organization to the application; thereafter, developers can choose on each Claude API request whether the customer or they themselves pay.

On data and compliance, Anthropic explicitly states that requests actually run within the developer's own organization, so the developer's rate limits, model access permissions, service tiers, data retention policies, and encryption keys all apply; arrangements on the customer's side such as zero data retention, business associate agreements, or customer-managed encryption keys do not automatically apply to these requests.

Related Articles

  • A bid for the largest IPO in human history, priced at 70–78 US dollars per share: Anthropic has been approved for listing under the ticker "ANT"
  • Claude API adds the ability to adjust effort: making AI work harder by task, shifting gears mid-way from casual chat to complex analysis
  • Anthropic upgrades its AI office suite: from brainstorming to data dashboards, Claude wants to take over the entire team collaboration workflow
  • Official fix underway: Claude Code reported to write passwords into project files, with 2,539 real credentials found on GitHub
  • The world's most powerful AI model: Claude Fable 5.2 arrives, surpassing GPT-6 Astra in coding and topping multiple benchmarks
  • Claude Code 2.1.311 adds a new environment variable, so AI agents can delete files without asking you

The translation is machine-generated and individual terms may be off. Nothing is rewritten or cut; the original is authoritative.

Published by IT之家
This site only summarises and aggregates; copyright stays with the original author. Headlines and summaries follow the source — where they differ, the original wins.
Read the original ↗