Claude Desktop
Experimental
This guide covers the Code tab in Claude Desktop. Chat and Cowork are outside its scope. For terminal setup, use the Claude Code guide.
You need a Divyam account and an API key. If you don't have them yet, follow Getting Started first.
Replace the shortened divyam-sw-v1-... shown in examples with your complete API key.
The examples use divyam-as/claude-opus-5-5. To choose another model, use this guide’s model selector when one is shown, or copy its current ID from Models and pricing. Check its context limit and supported features.
Switching between standard Claude Desktop and third-party inference changes which Code sessions appear, and existing sessions do not move into the new mode. Divyam has not verified that switching back restores them. Save important work and back up any session history you need before applying changes. See Claude Desktop's session storage guidance.
1. Open the Desktop connection form
Install a current Claude Desktop release using the official installation instructions.
Open Help → Troubleshooting → Enable Developer Mode, then Developer → Configure Third-Party Inference….
Desktop reads its own third-party inference configuration. The terminal's ANTHROPIC_BASE_URL and ANTHROPIC_AUTH_TOKEN exports do not set this connection. A managed policy may hide the form or make it read-only.
2. Add the Divyam connection
In Connection, enter:
| Setting | Value |
|---|---|
| Inference provider | Gateway |
| Credential kind | Static API key |
| Gateway base URL | https://api.preview.divyam.ai/anthropic |
| Gateway API key | Your complete Divyam API key, replacing divyam-sw-v1-... |
| Gateway auth scheme | Bearer |
Bearer sends the key in the Authorization header. x-api-key is the alternative scheme if you choose that authentication method.
3. Set the Desktop model list
Under Models, add this full ID to Model list:
| Model | ID |
|---|---|
| Opus | divyam-as/claude-opus-5-5 |
An explicit list of full model IDs avoids depending on automatic model discovery.
Choose Apply Changes, then Relaunch now. Open Code, choose a local project and select your model in Desktop's picker. Before the first message, choose Manual in the permission-mode selector so edits and commands ask for approval.
Hello, what are your capabilities?
Review the app's project and tool permission prompts as you work. The Desktop permission-mode reference describes the choices.
Check your requests
Open the Divyam console and select Logs. Your requests appear there within a few seconds, newest first, with their status, token counts and latency. Usage and Credits & Billing catch up within a few minutes.
Some clients display zero when model prices are missing. Check Credits & Billing for Divyam charges. Local dollar limits depend on the client's price and usage data.
For filters, cached-token counts and the requested-versus-routed view, see Logs and usage.