Guide

How to Use AtmoRouter with Cline in VS Code

Connect the Cline VS Code extension to AtmoRouter using an OpenAI-compatible endpoint, API key and model ID, with a practical troubleshooting checklist.

Cline can connect to AtmoRouter through its OpenAI-compatible provider setting. The connection needs three values: https://atmorouter.dev/v1, an AtmoRouter API key, and a complete model ID from the catalogue.

This configures the Cline extension inside VS Code. It does not configure VS Code's separate built-in chat experience, GitHub Copilot, or every other extension installed in the editor.

Prepare your account and project

Install Cline from the extension source linked by the official Cline documentation. Open its panel in VS Code.

Create an AtmoRouter key in API Keys. Name it for this client, such as cline-laptop, so you can identify its purpose later. Keep it in the extension's credential field rather than a source file that might be committed to Git.

Next, open Models and copy the exact ID you want to test. Check context length and whether the model fits an agent workflow. A model's ability to answer a chat prompt is not enough to establish reliable tool use.

Configure Cline's provider

The official Cline guide documents an OpenAI Compatible provider with base URL, key and model fields. Select that provider and enter:

  • Base URL: https://atmorouter.dev/v1
  • API key: your AtmoRouter key
  • Model ID: the exact catalogue ID, including the family prefix

If your Cline version exposes model capability settings, enter only capabilities supported by the selected model. Do not enable images, caching or other features simply because another model in the same family supports them. Labels can vary between extension releases; the linked provider guide is the reference for the current UI.

Run a small first task

Use a disposable repository for the first test. Ask Cline to identify its main files and propose one small change. Review the proposal before allowing it to write files or run commands.

After the request, open Usage logs. Confirm the model ID and status, then inspect the token counts, timing and cost. A coding task can contain several model requests, so compare the complete task's cost rather than just the first response.

If that test succeeds, try a representative task from your own workflow. A model that performs well on a short explanation may behave differently when the prompt contains a large repository context or several tool results.

Troubleshoot the connection in order

For authentication failures, verify that the credential belongs to AtmoRouter and has not been revoked. For model-not-found errors, copy the ID directly from the current catalogue instead of typing a familiar model name from memory.

For a missing route, make sure the base ends in /v1, without appending /chat/completions to the field. For tool execution failures, distinguish between a rejected API request and a local command that needs permission, a missing dependency or a valid working directory.

Use Errors & retries for API responses and Usage & request timing for measurements. Removing details from the prompt until a text-only request works is a useful way to isolate a feature mismatch.

Can the helper configure Cline automatically?

The AtmoRouter helper provides guided Cline instructions. Its guided mode is different from automatic configuration: you still enter the settings in Cline's UI. Check the helper's preview before applying any client configuration.

For another coding client, read OpenCode setup. For a terminal assistant, see installing Hermes Agent.

Newer

How to Install Hermes Agent and Connect It to AtmoRouter

Older

Sitemap Now Covers the Blog: 128 URLs, One Feed

Build with AtmoRouter.