Use Claude API and Workbench When Web App Is Down

It is a common scenario in modern cloud infrastructure: the Claude web application (claude.ai) goes completely offline, throwing gateway errors or stuck screens, while the underlying model API (api.anthropic.com) remains functional. This occurs because the web app relies on a complex stack of frontend CDNs, authentication servers (such as Auth0 or Clerk), and web hosting services that are independent of the core AI models.

When the web interface is down but you need to get work done, you do not have to wait for the web team to deploy a fix. You can bypass the broken web app entirely and access the models through alternative developer channels.

1. Verify the split outage Before changing your workflow, confirm that the API is indeed healthy while the web app is down: 1. Open your browser and navigate to the [Anthropic API Status Page](https://status.anthropic.com/). 2. Look closely at the breakdown. Check if "Web App" is experiencing a degraded state while "API" is marked as operational. 3. Try navigating to `console.anthropic.com`. If the Console loads successfully, the core developer APIs and auth servers are online, and you can proceed with the workarounds below.

2. Use the Anthropic Workbench as a web app alternative The Anthropic Developer Console features a built-in playground called the **Workbench**. It behaves similarly to the main chat interface, allowing you to select models, adjust system prompts, upload text, and generate responses using your API billing account. 1. Navigate to [console.anthropic.com](https://console.anthropic.com) and log in (this console uses a separate authentication flow from the main consumer web app). 2. Click on **Workbench** in the top navigation bar. 3. In the right-hand panel, select your desired model (e.g., `claude-3-5-sonnet`). 4. Type your prompt in the **User** block. 5. Click **Run** to generate a response. *Note: Using the Workbench incurs charges per token against your developer API credit balance, rather than using your Claude Pro subscription.*

3. Connect your API key to a local chat interface If you prefer a structured chat experience with sidebar history, you can use your Anthropic API key with an open-source, locally hosted interface. This gives you a permanent, outage-proof backup UI. 1. Go to the Anthropic Console, select **API Keys**, and generate a new secret key. Copy it securely. 2. Download or open a trusted third-party interface such as **LibreChat**, **TypingMind**, or **Big-AGI**. 3. Open the application settings, select **Anthropic** as the provider, and paste your API key. 4. You can now chat with Claude models directly through this alternative interface. Your prompt history will be saved locally on your machine, bypassing the Claude web app's databases entirely.

4. Clear Claude session cookies to resolve bad routing Sometimes, the web app is not actually down globally, but a corrupt local authentication cookie is preventing the frontend from communicating with the backend API. 1. On the `claude.ai` page, click the lock icon in your browser's address bar. 2. Select **Cookies and site data** or **Site settings**. 3. Click **Manage on-device site data** and delete all listed cookies and local storage items. 4. Close the browser tab, open a new one, and try to sign in again. This forces a clean handshake with the authentication servers.

When to escalate If both `claude.ai` and `console.anthropic.com` are completely inaccessible, or if you receive a `401 Unauthorized` error when attempting to run queries in the Workbench, the outage has spread to the authentication layer or core routing layers. In this case, local workarounds will not function. Monitor the official status page and wait for the engineering team to resolve the underlying system-wide network degradation.