Skip to content

Claude Desktop Integration Guide

What's Claude Desktop

Claude Desktop is Anthropic's desktop application for macOS and Windows. It gives you the full Claude chat experience in a native app, with file attachments, screenshots, and Model Context Protocol (MCP) tools.

How Claude Desktop Routes Through SIRAYA Model Router

Recent versions of Claude Desktop include a third-party inference option. When you select the Gateway (Anthropic-compatible) backend, the app sends its Messages API requests to whatever endpoint you configure instead of Anthropic's servers.

Because SIRAYA Model Router exposes an Anthropic SDK compatible API, you can point Claude Desktop at SIRAYA and run Claude — or any other model reachable through the router's Anthropic-compatible layer — directly inside the desktop app. This is the same routing you get in the Claude Code Integration Guide, configured through the GUI instead of environment variables.

Requirements

  • A recent build of Claude Desktop that includes Configure Third-Party Inference (under the Developer menu).
  • A SIRAYA Model Router account and API key.
  • The Gateway must speak the Anthropic Messages API (/v1/messages) — SIRAYA Model Router does.

Quick Start

1. Set up your SIRAYA Model Router account & API key

The first step is to create an account and get your API key.

2. Install or update Claude Desktop

Download the latest version from claude.com/download and install it. If you already have Claude Desktop, update it so the third-party inference option is available.

3. Enable Developer Mode

Open Claude Desktop, then go to Help → Troubleshooting and click Enable Developer Mode. This reveals the Developer menu used in the next step.

Enable Developer Mode under Help → Troubleshooting in Claude Desktop

4. Configure Third-Party Inference

From the menu bar, open Developer → Configure Third-Party Inference….

Developer → Configure Third-Party Inference in Claude Desktop

Set the backend to Gateway (Anthropic-compatible) and fill in the fields:

Field Value
Gateway Base URL https://llm.siraya.ai
Gateway API Key <API Key>
Gateway Auth Scheme bearer
Model List (optional) claude-sonnet-4.6, claude-opus-4.8, claude-haiku-4.5

Configure Third-Party Inference panel set to Gateway (Anthropic-compatible) with the SIRAYA base URL and API key

Do not add /v1 to the Base URL

Claude Desktop appends /v1/messages to the Gateway Base URL for you. Enter the root host (https://llm.siraya.ai) — not https://llm.siraya.ai/v1. If your first request fails with a path like /v1/v1/messages, the /v1 suffix is doubled.

Auth scheme

SIRAYA Model Router authenticates with a bearer token, so keep Gateway Auth Scheme set to bearer — the same credential the Claude Code guide passes as ANTHROPIC_AUTH_TOKEN. The Model List field is optional: leave it blank to auto-discover models, or pin specific SIRAYA model IDs to control what appears in the picker.

5. Apply and restart

Click Apply locally to save the configuration. Claude Desktop reads this configuration once at launch, so you must fully quit the app and reopen it for the change to take effect.

6. Continue with Gateway

On the startup screen, choose Or sign in with Gateway (shown as Local configuration). The model picker now lists the SIRAYA Model Router models you can run — pick one and start chatting.

Claude Desktop startup screen showing Continue with Gateway and the SIRAYA model picker

Choose the model you want to use. Note that Claude Desktop also shows "Claude" from the Model List field in the configuration or automatic detection

Claude Desktop chat screen with SIRAYA model

Field Reference

Setting Config key Value
Backend / Provider inferenceProvider gateway
Gateway Base URL inferenceGatewayBaseUrl https://llm.siraya.ai
Gateway API Key inferenceGatewayApiKey <API Key>
Gateway Auth Scheme inferenceGatewayAuthScheme bearer
Model List (optional) inferenceModels claude-sonnet-4.6, claude-opus-4.8, …

Troubleshooting

Continue with Gateway doesn't appear on startup

Claude Desktop only offers this option after a valid configuration is saved. Re-open Developer → Configure Third-Party Inference…, confirm the four fields, click Apply locally again, then fully quit and reopen the app.

Requests fail or models don't load

  • Verify your endpoint before configuring the app. From a terminal:
curl https://llm.siraya.ai/v1/messages \
  -H "Authorization: Bearer <API Key>" \
  -H "anthropic-version: 2023-06-01" \
  -H "content-type: application/json" \
  -d '{
    "model": "claude-sonnet-4.6",
    "max_tokens": 100,
    "messages": [{"role": "user", "content": "Hello"}]
  }'
  • Check the Gateway Base URL has no trailing /v1 (see the warning above).
  • Make sure you fully quit the app after Apply locally — a window reload is not enough.