Brief IA

Anthropic Restricts OpenClaw: End of Access Through Claude Subscriptions

🛠️ AI Tools·Tom Levy·

Anthropic Restricts OpenClaw: End of Access Through Claude Subscriptions

Anthropic Restricts OpenClaw: End of Access Through Claude Subscriptions
Key Takeaways
1Since April 4, 2026, OpenClaw and other third-party agents are excluded from Claude Pro and Max subscriptions by Anthropic.
2Users must now pay per use or utilize an API key, which can multiply costs by 10 to 50 times.
3Anthropic is offering a one-time credit equivalent to the monthly subscription until April 17, 2026, to facilitate the transition.
💡Why it mattersThis decision could disrupt the open-source ecosystem and affect independent developers who rely on these tools.
Le brief IA que lisent les pros

Le brief IA que les pros lisent chaque soir

Les 7 actus IA du jour, décryptées en 5 min. Gratuit.

Inclus dès l'inscription : notre sélection des meilleurs guides & comparatifs IA.

Choisis ton rythme

Gratuit · Pas de spam · Désabonnement en 1 clic

📄
Full Analysis

Claude subscriptions no longer cover third-party agents

Anthropic has decided to exclude third-party AI agent frameworks from its Claude Pro and Max subscriptions. As of April 4, 2026, users of tools like OpenClaw can no longer rely on their monthly plan to operate these agents. They must now go through a separate pay-per-use billing system or use an API key.

This announcement was made by Boris Cherny, head of Claude Code at Anthropic, who stated: “Claude subscriptions will no longer cover usage on third-party tools like OpenClaw. You can still use these tools with your Claude ID via additional usage bundles (now available at a discounted rate), or with a Claude API key,” he posted on X.

In practice, Pro subscribers who were using OpenClaw with their plan must choose between two options: purchase additional usage bundles, billed in addition to the subscription, with a discount of up to 30%, or use a standard API key, billed per token ($3 per million tokens for input and $15 for output for Sonnet 4.6, $15 and $75 respectively for Opus 4.6).

To ease the transition, Anthropic is granting a one-time credit equivalent to the monthly subscription amount for each user, usable until April 17, 2026. The company plans to extend this restriction to all third-party tools in the coming weeks.

A pricing shock for autonomous AI agent users

The gap between the actual cost of agent usage and the price of subscriptions largely explains this decision. Claude plans were designed for traditional conversational use: a chat exchange, a question, an answer. Frameworks like OpenClaw operate on a completely different model. A single instance running continuously (web browsing, calendar management, message processing, code execution) can consume the equivalent of $1,000 to $5,000 in API costs per day, depending on the load. Compared to a Max subscription at $200 per month, the imbalance is evident, Anthropic estimated.

Anthropic's subscriptions were not designed for the usage patterns of these third-party tools. “Capacity is a resource we manage carefully, and we prioritize our customers using our products and our API,” justified Boris Cherny.

For users, the shift to pay-per-use billing represents a cost increase estimated between 10 and 50 times the amount of the old plan for the most active profiles. Independent developers and hobbyists, who had structured their usage around the effective free nature of the subscription, are thus the most exposed.

Anthropic closes its ecosystem in response to the rise of OpenClaw

This measure comes amid a tense relationship between Anthropic and the open-source ecosystem. OpenClaw, the AI agent framework created by Austrian developer Peter Steinberger, had seen viral adoption in early 2026. However, on February 14, he left his own project to join OpenAI.

Anthropic's restriction, announced a few weeks after this departure, has sparked reactions. Peter Steinberger responded on X: “Dave Morin (a board member of OpenClaw, editor's note) and I tried to reason with Anthropic; the best we got was to delay it by a week.” Adding, mischievously: “Funny timing, first they copy popular features into their closed tool, then they lock down open source.”

The allusion directly targets Claude Cowork, Anthropic's AI agent platform launched in January 2026, which remains included in the Pro and Max subscriptions. Since then, Cowork has benefited from connectors, a plugin marketplace, and the deployment of "computer use" on Mac, reinforcing its positioning as a proprietary alternative to third-party frameworks. The message to developers is implicit: building within the Anthropic ecosystem remains covered by the plan, while building outside now costs the price of the API.

Brief IA — L'actualité IA en français

L'essentiel de l'actualité de l'intelligence artificielle, décrypté et expliqué chaque jour.