Anthropic reinstates OpenClaw and third-party agent usage on Claude subscriptions - with a catch

Key Points:

• Anthropic has reinstated the ability for Claude AI subscribers to use their subscriptions with third-party autonomous agents like OpenClaw, introducing a new "Agent SDK" credit system specifically for programmatic uses.
• This change reverses a prior April 2026 ban that prohibited such usage due to unsustainable token consumption and infrastructure strain, but now programmatic usage is limited to fixed monthly credits that do not roll over and are billed at API rates.
• Interactive Claude usage remains under standard subscription limits, while programmatic workflows, including third-party tools, draw from the new Agent SDK credits, creating a strict cap on usage without additional pay-as-you-go billing.
• The move aligns Anthropic’s subscription model with its API billing, ending the previous “compute arbitrage” where low-cost subscriptions were leveraged for high-volume agent use, but it has sparked significant backlash from developers who see it as a devaluation of their subscriptions.
• Despite criticism, this approach allows Anthropic to better manage compute resources and margins while still offering a sandbox for developers, making it a partial restoration for users reliant on autonomous AI agents like OpenClaw.