Anthropic’s Massive Account Bans: 3 Prohibited Practices You Must Stop Now

Developers who were using OpenCode overnight discovered their accounts disabled this morning; Anthropic’s recent crackdown applies equally to Pro and Max paid users.

1. Two‑step enforcement blocks third‑party access

On January 9 Anthropic deployed a new server‑side verification that rejects any request impersonating the official Claude Code identity, returning the same error message for all such calls:

"This credential is only authorized for use with Claude Code."

The first victims included OpenCode (over 100 k ★ on GitHub), OpenClaw, Cline, RooCode, and even xAI employees using Claude via Cursor. No advance notice was given before the disconnections.

On February 19 Anthropic codified the restriction in its service terms:

Free, Pro, Max subscription OAuth Tokens may not be used with any third‑party tool or Agent SDK; violations are treated as policy breaches.

Technical blocking preceded the contractual prohibition, and account bans have already occurred. A Reddit post titled “Anthropic is banning their $200/month heavy users” reported a Max subscriber whose account vanished the day after renewal. On January 27 a screenshot of a ban shared by developer Philipp Spiess spread quickly on X, sparking wide discussion.

2. Self‑check: Are you at risk?

The following three usage patterns are explicitly prohibited and can lead to account termination:

Using tools such as OpenCode, OpenClaw, Cline, RooCode that log in with a personal Claude account.

Configuring Pro/Max subscription OAuth Tokens or login credentials in any third‑party client.

Employing technical tricks (e.g., forged request headers) to bypass official Claude detection.

Key clarifications:

Bans are immediate and without prior warning.

Anthropic acknowledges accidental bans for users who unintentionally trigger the filter and says such accounts will be manually reinstated.

The enforcement focus is on heavy abusers—large token consumption and long‑running automated agents—while light users face lower risk.

Claims of “reporting the ban to authorities” are fabricated; BleepingComputer verified the screenshot is fake and Anthropic confirmed no such report was made.

3. Why the sudden tightening?

Claude Max costs $200 per month and includes rate limits that are manageable within the official Claude Code environment. Third‑party tools circumvented these limits, allowing automated agents to run continuously, turning a capped “$200 buffet” into effectively unlimited usage and causing Anthropic financial loss.

4. The OpenClaw example

Founder Peter Steinberger built OpenClaw in Vienna, amassing over 140 k GitHub stars and more than 1.5 million AI agents in 60 days. Anthropic responded with a trademark complaint, demanded a name change, and cut OAuth access. Consequently, the project migrated to OpenAI, with Sam Altman announcing Steinberger’s joining OpenAI on February 15 2026.

The crackdown shut down developers and drove ecosystem participants away.

5. What can you do now? Three viable routes

All solutions have been verified as of February 2026.

Route 1: Switch to an API Key and keep using your existing tool. Obtain an API Key from the Anthropic Console and pay per‑usage. OpenCode, Cline, RooCode and similar tools support API‑Key integration, preserving workflow while moving from subscription to metered billing. Suitable for heavy users who accept pay‑per‑call pricing.

Route 2: Use only official channels. Access Claude via the Claude.ai web UI or the official Claude Code CLI—these are the only compliant usage methods for Pro/Max accounts. Ideal for casual conversation or lightweight development without large‑scale agent tasks.

Route 3: Keep the tool, change the underlying model. After the ban, OpenCode was updated within hours to support OpenAI, Google Gemini, Groq, AWS Bedrock, OpenRouter and other providers. OpenAI has officially confirmed support for OpenCode users. By swapping Claude for another model, you retain your toolchain without incurring API fees.

Anthropic’s controls are expected to tighten further; this is unlikely to be the last enforcement action.