Breaking: Anthropic Reverses Ban on OpenClaw – Introduces Agent SDK Credits
San Francisco, April 14, 2026 – Anthropic has abruptly reversed its controversial prohibition on using Claude subscriptions to power third-party autonomous agents like OpenClaw, effective immediately. The company announced via its official developer account, @ClaudeDevs, that it is restructuring its paid subscription tiers to include a new allocation of “Agent SDK” credits specifically for programmatic use.
This marks a significant policy shift from the ban enacted just ten days ago, which barred subscribers from using their Claude Pro or Max plans to run external agentic harnesses. The reversal comes after intense backlash from the developer community, many of whom rely on open-source tools like OpenClaw for automated tasks.
“We heard our users loud and clear — they want flexibility,” said an Anthropic spokesperson in a statement. “The Agent SDK credits provide a dedicated path for third-party agent usage while ensuring our infrastructure remains sustainable.”
However, the restoration comes with a major catch: programmatic usage is now capped at a fixed, non-rollover monthly credit, valued at $20 to $200 depending on the plan, and billed at standard API rates. If users exhaust these credits, they cannot draw from their general subscription pool and must purchase additional usage credits.
Background: Why Anthropic Blocked OpenClaw in the First Place
On April 4, Anthropic explicitly prohibited the use of Claude subscriptions to power non-Anthropic agents, citing capacity and service issues. The problem was that some subscribers — paying just $20 to $200 per month — were consuming hundreds or even thousands of dollars worth of tokens through autonomous agents like OpenClaw.
This abuse of the flat-rate subscription model created unsustainable costs and strain on Anthropic’s inference infrastructure. While the ban technically blocked subscription-based access, users could still use OpenClaw by switching to the pay-as-you-go API or buying extra usage credits — a workaround that many found clunky and expensive.
“The previous policy was a blunt instrument,” noted Dr. Elena Torres, a professor of AI economics at Stanford. “Anthropic needed to contain runaway usage, but cutting off a popular open-source tool hurt developer trust. This new approach is more nuanced.”
What This Means for Claude Subscribers
Starting immediately, all paid Claude subscribers (Pro, Max, and Team tiers) will see a separate “Agent SDK” credit balance in their account dashboard. This credit is allocated per month and, crucially, does not roll over — unused credits expire at the end of the billing cycle. Usage of these credits is subject to the same API pricing structure that developers pay through the Claude API.
For power users who rely heavily on OpenClaw or similar third-party agents, this could mean a net increase in costs if their agent usage previously fell within the general subscription quota. Anthropic has not revealed how it will calculate the credit amount, but early reports suggest it mirrors the subscription fee — $20 for Pro, $100 for Max, and $200 for Team accounts.
“The real test will be whether these credits are sufficient for typical agent workflows,” said Marcus Chen, an open-source developer who contributes to OpenClaw. “Many agents consume tens of thousands of tokens per session. If your credit runs out mid-month, you’re stuck buying top-ups. It’s progress, but not a perfect fix.”
Anthropic emphasized that the Agent SDK credits are strictly for programmatic use — meaning they cannot be applied to Anthropic’s own tools like Claude Code or Claude Cowork, which remain covered by the general subscription. The company also noted that it will monitor usage patterns and may adjust credit allocations in future updates.
For now, users can reactivate OpenClaw on their subscription plans by authorizing the new “Agent SDK” integration in their account settings. Anthropic has published a step-by-step guide on its developer portal, including instructions for setting granular token limits to avoid exceeding the credit cap.
Agent SDK Credits: Key Details
- Non-rollover: Unused credits expire each month — no stacking or banking allowed.
- API rates apply: Usage is billed per million tokens, same as the commercial API.
- Separate pool: Cannot tap into general subscription usage limits once Agent SDK credits are exhausted.
- Manual opt-in required: Users must enable the feature; it is not automatically applied.
Anthropic’s move signals a broader trend of AI companies grappling with the tension between flat-rate subscriptions and variable compute costs. Competitors like OpenAI have introduced similar tiered credits for third-party tools, but Anthropic’s rapid policy reversal highlights the volatility in this space.