Skip to main content
Applies to: Claude Desktop 1.3883.0 (Updated: April 23, 2026)
Level: Beginner — no command line required, fully click-based, no Anthropic account needed

Introduction

Claude Desktop (enterprise version known as Claude Cowork) supports connecting to third-party LLM gateways, letting you use the client without an Anthropic subscription. This guide walks you through configuring AIHubMix as the API provider for Claude Desktop, enabling:
  • No Anthropic account required — bypasses phone number / email registration restrictions
  • Access to Claude, GPT, Gemini, DeepSeek, and more through a single AIHubMix key
  • Pay-as-you-go billing, flexible and cost-controlled

Prerequisites

Before you begin, make sure you have:
  • Operating System: macOS 13.0 (Ventura) or later, or Windows 10 / 11
  • AIHubMix Account: Register and top up at aihubmix.com

Step 1: Get Your AIHubMix API Key

  1. Visit aihubmix.com and log in
  2. Go to the console → API Keys page
  3. Click Create New Key, enter a name (e.g. claude-desktop)
  4. Copy the generated key (format: sk-xxxxxxxxxxxxxxxx)
Security notice: The API Key is only shown once. Save it somewhere safe immediately.

Step 2: Download and Install Claude Desktop

  1. Visit the official download page: claude.ai/download
  2. Download the installer for your operating system
macOS:
  • Open the downloaded .dmg file
  • Drag Claude.app into the Applications folder
  • Open Claude from Launchpad or the Applications folder
Windows:
  • Run the downloaded .msix installer
  • Follow the on-screen prompts to complete installation
  • Open Claude from the Start menu
Note: You do not need to log in to an Anthropic account after opening. Proceed directly to the next step.

Step 3: Enable Developer Mode and Configure AIHubMix

You need to enable Developer Mode before connecting a third-party API:
  1. Open the Claude Desktop app
  2. Click Help in the top menu bar
  3. Select Troubleshooting
  4. Click Enable Developer mode
    Image
  5. Once enabled, a new Developer menu will appear in the top menu bar.
Image
  1. Click Developer in the menu bar, then select Configure third-party inference
Image
  1. Fill in the following fields in the configuration dialog:
Connection — Gateway (Anthropic-compatible) Gateway Base URL — AIHubMix service endpoint
https://aihubmix.com
Gateway API Key — Your AIHubMix API Key
sk-xxxxxxxxxxxxxxxx
Image
  1. Click Apply locally → Relaunch
Img V3 02111 1bf57885 9883 46d1 81b1 4a1c4b45bb1g
  1. After restarting, select Continue with Gateway on the screen below
Img V3 02111 E41b4400 E68c 4c63 Ac88 434d2e69693g

Step 4: Configure Models and Verify

In Gateway mode, Claude Desktop constrains model names to Claude style — model names must start with claude-, so you cannot enter real model names like gpt, gemini, or deepseek directly, or the dropdown will have no usable models. The fix is model-name mapping: the client uses a claude--prefixed alias, and you map that alias to a real model on your AIHubMix Key. ① Configure the mapping on your AIHubMix Key Open the AIHubMix Key management page, create or edit a Key, and add “alias → real model” entries in the Model name mapping section:
claude-g-p-t-5.5 -> gpt-5.5
claude-gemi-3.1 -> gemini-3.1-pro-preview
claude-depsek-v4 -> deepseek-v4-flash
Configuring model name mapping in the AIHubMix dashboard
② Enter the aliases in Claude Desktop’s Model list Here you enter the pre-mapping claude- aliases — the model names Claude Desktop sends to AIHubMix, not the real upstream model names:
Claude Desktop Model list with the pre-mapping aliases
Once configured, these aliases appear in the Claude Desktop model dropdown for selection:
Claude Desktop model dropdown showing the configured alias models
Naming tips:
  • Use a claude- prefix for the Model ID.
  • Do not write real model family names like gpt, gemini, or deepseek directly; use aliases such as g-p-t, gemi, depsek.
  • The Model ID must match the left side of the AIHubMix mapping character for character, otherwise the request won’t hit the mapping.
For more — automatically switching to a backup model when the primary fails, billing by the final responding model, and so on — see Model Mapping and Fallback.
Note: The default model is Claude Opus 4.7, which has higher pricing and can lead to significant costs for everyday use. We recommend switching to another model. AIHubMix offers a wide range of cost-effective and free models — view the full list at aihubmix.com/models.
Once configured, restart Claude Desktop and verify:
  • The Cowork tab appears in the left navigation panel
  • You can send messages and receive responses normally
  • The model name shown in the bottom-right corner matches the model you configured
    Image
Congratulations! You’re now using Claude Desktop with your own configured models, powered by AIHubMix.

Troubleshooting

Error on Launch / Cannot Connect

Checklist:
  1. Confirm the connection type is set to gateway (not bedrock / vertex)
  2. Check that the Base URL is https://aihubmix.com (no trailing slash)
  3. Verify the API Key was copied in full (must start with sk-)
  4. Make sure your AIHubMix account has sufficient balance
View logs:
  • macOS: Open Console.app, filter by Claude
  • Windows: Open Event Viewer

Model Not Found Error

Make sure the model ID you entered is available on AIHubMix. You can query the available model list via API:
curl https://aihubmix.com/v1/models \
  -H "Authorization: Bearer sk-your-api-key"

Which Auth Scheme Should I Use?

AIHubMix uses the x-api-key authentication header. If you encounter a 401 error, try switching to bearer.

Advanced: Export Configuration File (Optional)

After completing setup, you can export the configuration for backup or team distribution: macOS — Click Export in the config UI to download a .mobileconfig file
Windows — Click Export in the config UI to download a .reg registry file


Last updated: April 23, 2026 | Based on Anthropic official docs + AIHubMix API documentation
Image