Anthropic changes Claude pricing for third-party and programmatic use

Anthropic is changing Claude subscription pricing so programmatic use will be billed with dedicated monthly credits at API rates, while interactive use through its own tools remains subsidized. The change follows a months-long rollout of blocks on third-party tools such as OpenCode and OpenClaw.

Anthropic changes Claude pricing for third-party and programmatic use

Anthropic is changing how Claude subscriptions work for programmatic use, ending the old subsidy for third-party tools and other automated workflows. According to the source article, the new model will give paid subscribers dedicated monthly credits for programmatic usage, billed at full API rates after opt-in, while use through Anthropic’s own interactive tools will remain subsidized.

The change is set to go live on June 15, with subscribers receiving an email on June 8 to activate the new credit system. The source says the pricing shift comes after a series of access restrictions that steadily narrowed third-party tools’ ability to use Claude subscriptions.

⚡ New to this?

This is a pricing and access change for Anthropic’s Claude subscriptions. Programmatic use means automated or tool-driven usage, while interactive use means a person working through Anthropic’s own interfaces. The news matters because it changes what developers can do under a subscription, especially if they use third-party tools or self-built automation.

🦞 OpenClaw angle

If your workflows depend on claude -p, the Agent SDK, or third-party harnesses, assume they will be billed at API rates from here on out. Recalculate the cost of any agent loops, CI jobs, or scheduled automations before June 15 and move anything expensive to a usage-capped queue. If you build self-hosted tooling around model access, design for provider pricing changes and avoid coupling your system to subscription-only behavior.

On January 9, 2026, Claude Code deployed server-side checks that blocked third-party tools from authenticating to Claude subscriptions via OAuth, according to the article. OpenCode, an open-source terminal coding assistant, stopped working, and users were met with an error message. There was no prior announcement, the source says.

The policy then moved into formal language. On February 19, Anthropic updated its Terms of Service to explicitly ban the practice. On April 4, OpenClaw and other third-party harnesses were cut off from subscription access entirely, the article says. Boris Cherny, head of Claude Code, attributed the restrictions to “unsustainable demand.”

Anthropic’s latest explanation is that the new credit system simplifies billing. The source argues that the pattern over the last five months points to something different: third-party tools are being pushed out of the subscription subsidy, while Anthropic’s own surfaces remain privileged.

The announcement drew criticism from several developers and builders. Jeremy Howard said the company’s use of the word “interactive” was misleading because it appears to mean “using an Anthropic front-end,” not simply a human talking back and forth with the model. In his view, a person using claude -p or the Agent SDK interactively is now treated as programmatic usage.

The source article says the distinction matters because the same work can now be priced differently depending on where it happens. A person using Claude through Anthropic’s official interfaces may still be covered by the subscription subsidy, while the same token usage through a third-party harness can be billed at API rates.

The article also points to Anthropic’s privacy documentation, which says consumer plan data can be used for model training if Model Improvement is enabled. It says chat and coding session data can include the entire conversation, custom styles, conversation preferences, and data collected when using Claude for Chrome.

Peter Steinberger, creator of OpenClaw, argued that Anthropic had been absorbing features from third-party tools into Claude Code and then shutting out alternatives. The article lays out a timeline it says supports that view: January 9 for the first server-side blocks, late March for policy warnings about exceeding limits, April 4 for billing enforcement, and May 13 for the official pricing split.

Other developers framed the change as a straightforward price increase for heavy users. Reuven Cohen, an Agentic Engineer at Cognitum One, said the old $200 plan had “fuzzy limits” but high value for programmatic use, including recursive agent loops, headless automation, CI/CD pipelines, and multi-agent swarms. He described the new credit system as “a rate limiter wearing a party hat.”

Theo, another developer quoted in the article, criticized Anthropic for focusing on blocking better apps and harnesses instead of improving Claude Code itself. Some users, the source says, are already moving to OpenAI’s Codex, DeepSeek, or Kimi.

The article says the pricing change lands in a broader commercial context. José Valim, creator of Elixir, argued that the announcement was aimed at investors and enterprise customers rather than developers. Anthropic is reportedly seeking new funding at a $950 billion valuation, and the company has said its revenue is growing 80x annualized.

The source’s bottom line is that the subscription subsidy for programmatic use is ending. For developers who built automation, agent loops, and tooling around Claude subscriptions, the new pricing starts June 15.

Source: Kilo Blog ↗

More from Security News