VS Code + Claude Code

There are three common ways to use AI coding tools with Tokeness inside VS Code:

Approach	Reference
Run Claude Code in the VS Code terminal	See the Claude Code page
Use the Cline extension	See the Cline page
Use another extension that supports a custom Base URL	Follow the general fields on this page

Tokeness exposes a general OpenAI-compatible endpoint:

https://n.tokeness.dev/v1

1. Run Claude Code in the VS Code terminal

If you only use VS Code as a terminal, the configuration is the same as for regular Claude Code:

1. Install Claude Code.
2. Install CC Switch.
3. Add the Tokeness provider for Claude Code in CC Switch.
4. Enable routing mode.
5. Run claude in the VS Code terminal.

In this scenario, do not re-enter the key in the VS Code extension settings. Claude Code reads its own configuration.

See: Claude Code

2. Use the Cline extension

Cline is a VS Code extension that supports an OpenAI Compatible provider.

Field values:

Provider: OpenAI Compatible
Base URL: https://n.tokeness.dev/v1
API Key: your Tokeness API key
Model ID: copied from the Tokeness model marketplace

See: Cline

3. Use other VS Code AI extensions

As long as an extension exposes these fields, you can try connecting it to Tokeness:

Extension field	Tokeness value
API Provider	OpenAI Compatible / OpenAI
API Key	Tokeness API key
Base URL / API Base / Endpoint	https://n.tokeness.dev/v1
Model / Model ID	Full model name from the Tokeness model marketplace

If an extension only supports logging in to a specific vendor and does not allow a custom Base URL, it cannot connect to Tokeness directly.

4. Claude Code LLM Gateway notes

Claude Code officially supports connecting to an LLM Gateway via ANTHROPIC_BASE_URL, but this requires the gateway to expose the Anthropic Messages format.

The Claude Code integration in the Tokeness documentation defaults to the CC Switch routing mode, because it can redirect Claude Code requests to an OpenAI-compatible provider. Only when you explicitly use an Anthropic Messages-compatible endpoint should you follow the Anthropic API page.

Related pages:

• Claude Code
• Anthropic API

5. Keys and quota

VS Code extensions typically read project context, so their consumption is higher than a plain chat session.

Recommended practices:

• Create a separate key for each VS Code extension.
• Set a quota on the key.
• Start with read-only tasks on first use.
• Confirm the Git working tree status before making large-scale changes.
• Filter usage logs by key to review call records.

6. Verification

In the extension or Claude Code, enter:

Please reply with a single sentence confirming the VS Code integration is working.

Then check the Tokeness usage logs:

1. A new request appears.
2. The key is the one dedicated to the VS Code extension.
3. The model name is correct.
4. The status code is successful.

7. Troubleshooting

Symptom	Fix
Extension does not show a Base URL field	The extension may not support custom APIs
Authentication failed	Check the key, whitespace, and whether it is enabled
404 Not Found	Set Base URL to https://n.tokeness.dev/v1
model not found	Re-copy the model name from the model marketplace
No Claude Code logs	Check whether CC Switch has Tokeness enabled and routing mode on
Extension can chat but cannot edit code	Switch to a model with stronger coding ability, or use Cline / Claude Code
Abnormal consumption	Limit the context range and check for looping tasks

External docs

• Claude Code LLM Gateway
• Claude Code Settings
• Cline OpenAI Compatible Provider