Anthropic restricts Claude subscriptions to curb third-party tool usage amid service strain

Anthropic has moved to restrict how its Claude AI service can be used, specifically banning subscription-based access through third-party tools like OpenClaw. The decision comes as the company grapples with degraded service and struggles to balance capacity with surging demand for its AI models.

The restriction, announced on April 6, 2026, prohibits Claude subscribers from using their accounts with external tools that harness the AI service through subscriptions. This policy shift follows a period of service instability, with Claude.ai experiencing a partial outage on the same day the announcement was made.

Service disruptions and capacity challenges

Anthropic's status page reported elevated errors on Claude.ai, including desktop and mobile platforms, characterizing the incident as "a partial outage." The company's uptime over the past 90 days had slipped to 98.82 percent. The disruption, which affected login functionality and some product features including voice mode, lasted from 15:00–16:30 UTC on April 6.

While Anthropic did not directly attribute the service issues to third-party tool usage, the timing of the restriction announcement suggests a connection. The company has been working to optimize its systems for specific workloads and manage capacity thoughtfully as demand for Claude continues to grow.

The economics of subscription vs. API usage

The restriction highlights a significant pricing disparity between Claude's subscription model and API usage. Subscribers pay a flat rate with session and monthly usage limits, while API customers pay per token with no usage caps. For heavy users, subscriptions can offer substantial savings - one reporter reported using approximately $236 worth of tokens for a $20 monthly subscription in March.

This pricing advantage has led many developers to use third-party harnesses like OpenClaw to interact with Claude Code, maximizing the value of their subscriptions. However, this usage pattern puts "an outsized strain on our systems," according to Anthropic, as subscriptions weren't designed for the consumption patterns of these external tools.

Industry context and enforcement

Anthropic's move mirrors similar actions taken by other AI providers. In February, Google enforced its terms of service regarding third-party use of its Gemini CLI and related services, citing violations of applicable terms and policies.

The restriction follows earlier policy warnings from Anthropic. In February, the company reaffirmed its existing policy forbidding third-party harnesses with Claude subscriptions, around the same time OpenClaw began gaining attention as an autonomous AI agent platform.

Mitigation and alternatives

To soften the impact of the sudden policy change, Anthropic offered several accommodations:

• One month of extra usage credit based on subscribers' monthly plans
• Extra usage bundles at 30 percent off
• Option to cancel plans with refunds
• Continued ability to use Claude with third-party tools through extra usage bundles or API keys

The company emphasized that Claude subscriptions remain valid for its core products: Claude.ai, Claude Code, and Cowork.

Broader implications

As Anthropic works toward going public, the company faces pressure to ensure its customer acquisition strategy doesn't lead customers toward rival products or magnify costs. The restriction on third-party tool usage represents a significant shift in how the company manages its resources and customer relationships.

The move also raises questions about the sustainability of current AI pricing models as demand continues to grow. With more than half of US datacenters planned to open this year facing delays, capacity constraints may continue to be a challenge for AI providers.

For developers and businesses relying on AI tools, the restriction underscores the importance of understanding service terms and the potential for sudden policy changes as providers work to balance growth with operational stability.