Using aicentos with Claude Desktop

Scope

This guide is for the new Cowork / Code third-party inference mode in Claude Desktop, which Anthropic documents as third-party inference.

Use this setup if you want to:

• use Cowork in Claude Desktop
• use Code in Claude Desktop
• route requests through aicentos via a custom API or gateway

Note

This is not the old "just sign in and chat" flow. With third-party inference enabled, you mainly use the Cowork and Code tabs.

Prerequisites

Prepare the following first:

1. The latest Claude Desktop
2. A aicentos token
3. A token that can access Claude models

Get your token here:

• aicentos

Recommended starter models:

• claude-sonnet-4-5-20250929
• claude-opus-4-5-20251101
• claude-haiku-4-5-20251001

Method 1: Configure Directly in Claude Desktop (Recommended)

Anthropic recommends using the built-in third-party inference setup flow in the desktop app.

1. Enable Developer mode

In Claude Desktop, go to:

Help → Troubleshooting → Enable Developer mode

After enabling it, you should see Developer menu entries.

2. Open third-party inference setup

Then go to:

Developer → Configure third-party inference

This opens Anthropic's official third-party inference configuration flow.

3. Choose Gateway mode

Select:

Gateway

aicentos maps to the gateway mode in this flow.

4. Fill in the aicentos settings

Use the following values:

• Gateway URL: https://www.aicentos.com/
• Authentication: x-api-key
• API Key: your aicentos token

If the UI asks for models, start with:

claude-sonnet-4-5-20250929
claude-opus-4-5-20251101
claude-haiku-4-5-20251001

If there is an option to enable Code, keep it enabled.

5. Apply locally

When finished, choose:

Apply locally

This writes the configuration into your local Claude Desktop user config.

After that, restart Claude Desktop. In most cases you should then see the Cowork / Code workspace.

Method 2: Manage It via Local Config File (Advanced)

If you want to back up, migrate, or troubleshoot the setup, inspect Claude Desktop's local config file.

Common locations:

macOS

~/Library/Application Support/Claude-3p/claude_desktop_config.json

Windows

%APPDATA%\Claude-3p\claude_desktop_config.json

Linux

~/.config/Claude-3p/claude_desktop_config.json

Tip

The local file stores the third-party inference data under enterpriseConfig. The safer path is to configure it in the app first, then back up the generated config rather than hand-writing JSON from scratch.

Recommended aicentos Values

If you want to verify the key fields manually, focus on these:

Field	Recommended Value
inferenceProvider	gateway
inferenceGatewayBaseUrl	https://www.aicentos.com/
inferenceGatewayAuthScheme	x-api-key
inferenceGatewayApiKey	your aicentos token
inferenceModels	list of Claude models
isClaudeCodeForDesktopEnabled	true

Recommended model list:

claude-sonnet-4-5-20250929
claude-opus-4-5-20251101
claude-haiku-4-5-20251001

Usage

Once configured, you can use Claude Desktop to:

• open Cowork for general tasks, docs, and file collaboration
• open Code for coding, terminal workflows, and engineering tasks

If the model picker is visible, start with:

• daily use: claude-sonnet-4-5-20250929
• deep reasoning: claude-opus-4-5-20251101
• fast lightweight tasks: claude-haiku-4-5-20251001

FAQ

Why can't I find Enable Developer mode?

Check these first:

1. You are on a recent Claude Desktop build
2. You are looking at the real system menu bar, not just the window controls

If it still does not appear after upgrading, fully quit Claude Desktop and reopen it.

What if Windows shows Virtual Machine Platform not available?

If Claude Desktop shows:

Claude's workspace requires the Virtual Machine Platform on Windows. Enable this feature, then restart.

It means Virtual Machine Platform is not enabled on Windows. Open PowerShell as Administrator and run:

dism.exe /online /enable-feature /featurename:VirtualMachinePlatform /all /norestart

After the command completes, restart Windows and reopen Claude Desktop.

Why can't I see Cowork / Code after configuring it?

Check in this order:

1. Developer mode is enabled
2. Configure third-party inference was completed
3. You selected Gateway
4. gatewayUrl is https://www.aicentos.com/
5. The token is valid and can access Claude models
6. Claude Desktop was fully restarted after setup

Why is the Gateway URL not https://www.aicentos.com/v1?

Because Claude Desktop third-party inference uses Anthropic-style gateway configuration, not the OpenAI-compatible /v1/chat/completions path. Use the gateway root URL instead:

https://www.aicentos.com/

What if authentication fails after setup?

Check these first:

1. Authentication is set to x-api-key
2. The API key field contains your aicentos token directly
3. The token is still valid and has the required permissions

Can I edit the local JSON manually?

Yes, but it is not the best choice for the initial setup. Claude Desktop evolves its third-party inference fields over time, and the in-app flow is more stable and easier to recover from if something breaks.