Anthropic halts OpenClaw access for Claude subscriptions as part of a cost-cutting initiative | TNW

      In summary: Anthropic has disabled the ability for Claude Pro and Max subscribers to utilize flat-rate plans with third-party AI agent frameworks, starting with OpenClaw, effective April 4, 2026. This shift places the financial burden of running autonomous agents on users through a pay-as-you-go billing model. The creator of OpenClaw, who joined OpenAI in February, criticized the decision as a betrayal of open-source developers. As a result, thousands of users are facing increases in costs of up to 50 times their previous monthly expenses.

      Anthropic has ceased a discreet subsidy that positioned its Claude models as the preferred choice among the open-source AI agent community. As of April 4, 2026, users on Claude's Pro and Max subscription tiers are prohibited from directing their plan's usage limits through third-party tools like OpenClaw. To continue utilizing these tools with Claude, they will need to make separate payments under a new “extra usage” billing system. Anthropic has indicated that this restriction will soon extend to all third-party frameworks.

      The announcement shocked many developers who had designed their personal AI systems under the assumption that a flat monthly subscription would suffice. Now, this is no longer the case for many.

      The rationale behind the change is clear, though the timing was unexpected. Claude’s subscription plans were geared towards conversational interactions: a user initiates a chat, asks a question, and receives a reply. However, agentic frameworks operate on a different model altogether. A single OpenClaw instance running continuously for a day, performing tasks such as web browsing, calendar management, responding to messages, and executing code, can incur API costs between $1,000 to $5,000, depending on the workload. This presents an unsustainable transfer of computational costs from the user to Anthropic, given the $200-per-month Max subscription.

      “Anthropic’s subscriptions were not designed for the usage patterns of these third-party tools,” stated Boris Cherny, Head of Claude Code at Anthropic. “We manage our capacity thoughtfully and are prioritizing our customers who use our products and API.”

      The scale of the issue was considerable. At the time of the announcement, it was estimated that over 135,000 OpenClaw instances were operational, and analysts noted a price disparity of more than five times between what significant agentic users paid under flat subscriptions compared to API costs for equivalent usage. Essentially, Anthropic’s subscription model had been quietly subsidizing usage that it had not properly priced.

      What OpenClaw is and its significance

      OpenClaw is an open-source AI agent framework developed by Austrian programmer Peter Steinberger. It was initially launched in November 2025 as Clawdbot, a side project aimed at exploring the outcomes of providing a large language model with persistent memory, tool access, and the capacity to communicate via messaging platforms like WhatsApp and Telegram. The outcome revealed a strong demand for such capabilities.

      The project underwent two renames within three days in late January 2026: first to Moltbot due to trademark issues raised by Anthropic over its similarity to “Claude,” and then to OpenClaw shortly thereafter. By March 2, 2026, the repository had amassed 247,000 stars and 47,700 forks on GitHub, becoming what many analysts regarded as the fastest-growing GitHub project ever, reaching 100,000 stars in less than 48 hours at its peak. The framework supports over 50 integrations and is compatible with Claude, GPT-4o, Gemini, and DeepSeek. Tencent even developed an enterprise platform based on it, showcasing OpenClaw’s significant influence beyond individual enthusiasts.

      A timing dilemma

      The restriction takes on greater significance in light of events from February. On February 14, 2026, Steinberger announced his departure from his project to join OpenAI, where Sam Altman stated he would “drive the next generation of personal agents” and that OpenClaw would transition to an open-source foundation with ongoing support from OpenAI. Steinberger noted that “partnering with OpenAI is the fastest way to make this available to everyone.”

      Anthropic’s restrictions were implemented shortly after this announcement, a timeline that has not gone unnoticed. Steinberger and investor Dave Morin sought to negotiate a smoother transition by approaching Anthropic, but they could only delay enforcement for one week.

      “First, they copy some popular features into their closed framework, then they lock out open source,” Steinberger responded regarding the ban.

      Whether this timing reflects a strategic move or mere coincidence, the outcome remains the same: the most popular open-source agent framework, now linked to OpenAI, has effectively been priced out of Claude’s subscription offerings.

      The cost impact on users

      For developers accustomed to unlimited agentic runs on a flat fee, the new billing structure presents a significant disruption. Under the pay-as-you-go extra usage model, per-interaction costs are estimated between $0.50 to