Just now, a "closed-door" incident occurred in the AI circle that can be recorded in history. Anthropic has officially banned the use of its own packages to access OpenClaw! Boris Cherny, the father of Claude Code, announced that starting from 3:00 pm Eastern Time on April 4th (3:00am Beijing time on April 5th), Claude will block all third-party tools and can only use additional packages or APIs to use these tools.
This means that thousands of developers and start-up teams who rely on OpenClaw to improve efficiency have lost the "unlimited" usage bonus overnight and were forced to switch to extremely expensive"Pay as you go"model.
A loyal user of Claude on OpenClaw received a heavy blow.
The timing of this news announcement is also very intriguing - Peter Steinberger, the father of OpenClaw, just switched to OpenAI not long ago, and what Anthropic wants to do is clearly revealed!
It can be said that this is a commercial revenge disguised as a policy.
This news quickly topped the list of Hacker News in the developer community.
Let us remember this day: April 4, 2026. From this day on, the AI industry has moved from open collaboration to giant separatism.
Claude's official email proves that this time, Anthropic is using OpenClaw for surgery!
This policy will be enforced first on OpenClaw starting April 4th, but applies to all third-party toolchains, and will be rolled out to more tools soon.
As an appeasement, Anthropic gave a one-time subsidy equal to one month's subscription fee. Valid for collection before April 17th.
The hammer falls
Note that Anthropic’s email was sent at a very shameful time.
Friday evening, April 3, is the favorite time for Internet companies to send out bad news.
Anthropic issued a notice to thousands of OpenClaw users: starting tomorrow, your Claude subscription quota can no longer be used for OpenClaw. If you want to continue, then you can pay according to your usage.
https://x.com/VadimStrizheus/status/2040199979927482618
After three months of siege and suppression, the last blow was finally delivered.
The trigger: a "defection" that could change the situation
Why would Anthropic kill an open source tool at this point in time, regardless of the consequences?
Because Peter Steinberger, the soul of OpenClaw and the father of lobster, joined their "rival" OpenAI.
At one time, Peter Steinberger was one of the developers who understood the Claude ecosystem best. His OpenClaw made Claude extremely easy to use.
Now, for Anthropic, OpenClaw has become a Trojan horse for the enemy.
Anthropic believes that OpenClaw is no longer a pure efficiency tool, but an "intelligence collector" extending into its own back garden.
Now that the founder has become an OpenAI person, your tools should no longer take advantage of my subscription quota.
Peter himself had no choice but to speak out, suggesting that Anthropic was "closing the door and beating the dog" and "prostituted" the open source community for free:
Dave Morin (OpenClaw board member) and I tried to persuade Anthropic to stay calm.
But in the end, all we could achieve was to delay the arrival of this day by a week.
https://x.com/steipete/status/2040209434019082522
Developers are in mourning
Budget explodes overnight
For ordinary developers, this ban is simply a "dimensionality reduction blow."
Previously, many developers achieved extremely low-cost automated workflows by purchasing Claude's fixed monthly subscription and using OpenClaw's powerful interface.
Spending $20 to buy a Claude Pro package allows the lobster to call Claude to work 24 hours a day, 7 days a week. If you use the API channel for the same usage, the bill may run into thousands of dollars.
One is a Max subscription with a cap of $200, and the other is a four-digit API fee.
Now, this road has been cut off by Anthropic himself.
Pay-as-you-go billing means that it is no longer a monthly package, and the cost is extremely uncontrollable. The AI budgets of many small and medium-sized teams were originally locked on a monthly basis, but now they may explode at any time.
What's even more cruel is that if you don't want to pay this sky-high toll, you must reluctantly reconstruct the entire business logic within 24 hours.
One word - absolutely!
Lobster stirs up the world, and grudges have already begun
The feud between the Father of Lobster and Anthropic started a long time ago.
Steinberger once complained publicly that Anthropic’s dealings with him “basically relied on legal letters.”
The first move,brand cutting.
At the end of January, a lawyer’s letter forced Clawdbot to change its name.
The second move,technical blockade.
On January 9, Anthropic quietly implemented a test on the server side: if the subscription token is not issued from the official Claude Code client, it will be directly rejected.
The core gameplay of OpenClaw was reset to zero overnight.
The third move,Qualitative terms.
In mid-February, the terms of service were updated: it is illegal to use the OAuth tokens of Free, Pro, and Max accounts in any third-party tools.
The fourth most ruthless trick is functional bargain hunting.
Claude Cowork launched Dispatch, which allows mobile phones to remotely control Claude on the desktop to work; Claude Code launched Channels to connect Telegram and Discord.
Within four weeks, the core functions of OpenClaw were officially replicated one-to-one!
In the words of AI blogger Matthew Berman: "They directly built OpenClaw themselves."
Technology media Semafor reported this morning that Anthropic chief commercial officer Paul Smith was asked whether customers were asking the company to build its own OpenClaw, and he admitted that this was indeed the case.
https://www.semafor.com/article/04/03/2026/anthropic-eyes-its-own-version-of-openclaw
And today, this email on April 4th is the final blow.
Yang conspiracy: forcefully promote one’s own “son”
Claude Cowork
While banning OpenClaw, Anthropic also took advantage of it.
They started hinting wildly:Don’t use those unreliable third parties, come and try our native Claude Cowork!
Claude Cowork allows Claude to have deeper control over the coding environment and computer interface.
In fact, this is not Anthropic’s exclusive gameplay.
This is what the giants do best”platform lock"Strategy:
first step:Use third-party open source tools to attract developers and expand the ecosystem.
Step two:Find excuses (such as security, infrastructure pressure, etc.) to block third parties.
Step three:Forcing users to migrate to their own more expensive and strictly controlled native integration tools.
This typical "vertical integration" approach allows them to firmly control the entrance and experience in their own hands, while at the same time allowing those "uncontrolled connectors" in the ecosystem to gradually lose their advantage.
Interestingly, OpenAI chose a completely opposite path at almost the same period.
OpenAI explicitly allows Codex subscriptions to be used in third-party clients such as OpenClaw.
In March, it went a step further and announced that it would provide free ChatGPT Pro access to open source project maintainers, and OpenClaw was named on the beneficiary list. OpenAI has taken in all the people Anthropic wanted to get rid of!
No one is completely innocent
To be fair, Anthropic has its reasons.
Every user who spends $200 on a subscription and uses APIs worth a thousand dollars is making the company lose money.
Third-party tools bypass official telemetry, disguise client identities, and create monitoring blind spots. These are real engineering and security risks.
What's more, Steinberger is now sitting in the OpenAI office - OpenClaw, in Anthropic's eyes, has changed from "a somewhat annoying diner" to "a spy sent from the opposite side."
Moreover, OpenClaw itself is not clean.
It has been exposed to a high-severity vulnerability (CVE-2026-25253) with a CVSS score of 8.8. An attacker can steal the user's authentication token through a link.
https://nvd.nist.gov/vuln/detail/CVE-2026-25253
On the public Internet, security agencies have discovered more than 30,000 OpenClaw instances with open doors.
But these rationalities cannot dispel the feeling of being cheated by developers.
The platform first opened the door, waited for people to come in and built the building, and then announced other arrangements - this script has been played out too many times.
X kills third-party clients, Apple tightens the App Store, and Google cuts free APIs. It’s the same story.
Every time, it’s the developers who get hurt.
Open ecology enters dusk
AI giants are closing themselves off to the country
Today Anthropic’s ultimate trick makes us understand a heartbreaking fact——
The golden age of free and open AI that belonged to developers is coming to an end.
Once, we thought that AI would be like the early Internet, based on protocol sharing and community evolution. But the reality is that the model owner holds the power of life and death, and can undo the efforts of thousands of people at any time.
Now, the countdown to the 3pm ultimatum on April 4 has begun.
Developers face three cruel choices——
Should I swallow the bitter pill, endure the sky-high pay-as-you-go billing, and continue using OpenClaw until the budget is exhausted?
Or should we migrate entirely, kneel down under Anthropic’s native tools, and accept platform lock-in?
Or, leave the scene angrily and completely abandon Claude.
Finally, when the giants are busy fighting with gods, please don’t forget that it is thousands of developers who have used lines of code to build your throne worth hundreds of billions.
This time, Anthropic won the competition, but completely lost the trust of the community!