Configure Multiple Models in Cline

Cline is a VS Code coding assistant that can read projects, edit files, and run commands. You can switch between Claude, GPT, Gemini, and Grok in one extension, with model requests routed through Passion8. Prepare your dashboard API key and an enabled model ID before installing and configuring the extension.

#Choose the Provider and URL

The provider determines the request protocol; the model ID selects the model. A GPT name does not guarantee Chat Completions support, and a Gemini name does not require the Google provider.

Enter your Passion8 API key for all four configurations. The example IDs illustrate the format; availability depends on your token permissions and the current route.

#1. Install the Extension

Search for Cline in VS Code's Extensions view, verify the official listing, and install it. Open your project folder, open Cline from the Activity Bar, and click its settings icon to configure the API.

If the code command is available, you can install from a terminal:

code --install-extension saoudrizwan.claude-dev

Choose the option to enter your own API provider and key. This tutorial connects directly to Passion8 and does not require creating a vendor account.

#2. Configure Claude

Enter these settings in Cline:

Anthropic API Key is the extension's field label; enter your Passion8 key here. Enable the custom URL or requests will go to the default Anthropic address.

The native Anthropic path helps preserve Claude's message, tool, and caching semantics. If Cline's Model dropdown does not include a new dashboard model, update the extension first. If it is still missing, do not copy an older model's capability settings to the new one. Choose a Claude model supported by both your account and the extension, or explicitly configure the model in the official Claude Code extension.

#3. Configure GPT, Gemini, or Grok

These three use the same settings; change the model ID:

If the required ID is missing from the list, choose Use custom model ID… or the equivalent manual entry field in your version. A failed model-list refresh does not necessarily mean inference will fail: enter the exact ID and verify with a small request.

Keep advanced options at their defaults for the first connection. Do not enter an Azure API version or enable Azure Identity Authentication, and do not add official account credentials. Adjust parameters only when the selected model explicitly requires them.

#Choosing the Native Gemini Provider

This tutorial recommends the OpenAI Compatible path for Gemini. Cline's official Gemini configuration page does not describe a consistent custom URL setup. Some versions support a Gemini Base URL internally, but the settings interface varies by version.

Use the native path only if your installed version explicitly exposes a custom Gemini Base URL and accepts the exact dashboard model ID. Select Google Gemini, set the custom root to https://passion8.cc, and enter your Passion8 key. This path requests Gemini /v1beta/models/... endpoints; do not enter https://passion8.cc/v1. If you cannot find the custom URL setting, continue with OpenAI Compatible.

#4. Check Plan and Act

Cline's Plan mode discusses an approach and reads context; Act executes agreed changes. Settings can allow the two modes to use separate providers and models.

For your first setup, use the same Passion8 URL, key, and model in both modes. If separate configuration is enabled, check both Plan and Act. Updating only one mode can leave the other using an old provider.

Later, you can use a fast model to clarify requirements and another model for complex edits. Switching vendors changes tool, reasoning, and context formats. Prefer starting a new task after switching models, with the agreed plan saved in a project file or included in your next prompt.

#5. Reload and Verify

Save the settings, run Developer: Reload Window in VS Code, reopen Cline, and create a new task. First send:

Reply with one sentence saying "Connected successfully". Do not read files, edit files, or run commands.

After receiving a response, check the matching request time, model, and usage in the Passion8 dashboard. A model's self-reported identity does not establish the route; dashboard records are more direct.

Then verify project access:

Read only package.json and describe the framework and test commands. Do not edit files or run commands.

Finally, request a small, clearly scoped change: ask for a plan first, limit edits to specified files, run existing relevant checks, and review the diff. Successful text chat does not establish compatibility with tools and streaming. File reading and a small edit help verify those steps separately.

#Project rules and task context

Create .clinerules/project.md at the project root and record build commands, coding conventions and acceptance criteria. Current Cline supports both .clinerules/ and .cline/rules/; choose one. The VS Code Rules panel creates workspace rules in .clinerules/ by default.

# Project rules

- Read the existing implementation before changing it.
- Preserve unrelated local changes.
- Use the package manager specified by the lockfile.
- Keep each change within the requested file scope.
- Run the relevant existing checks and report the actual results.

Confirm that the rules are enabled in Cline's Rules panel. Add the project's real start and test commands, rather than copying commands that do not exist. Keep keys out of rule files; the same project rules can be used with different models.

In a new task, reference the files you want to work on, state the goal and acceptance criteria, then ask the model to read the implementation. For large projects, search first and read relevant files instead of sending the whole repository each time. Do not assume Grok Build's AGENTS.md or Gemini CLI's GEMINI.md is the same rules entry point as Cline's.

#Capability and Cost Settings

Context Window Size, Max Output Tokens, image capabilities, and prices under Model Configuration help Cline manage tasks and estimate costs. They do not expand the upstream model's capabilities or change Passion8 billing.

Automatic approval for reads, writes, and terminal commands uses separate permissions. Keep manual approval initially, then adjust after confirming tool behavior and the scope of edits.

#Troubleshooting

#Configuration Questions

#Is Cline an official Claude, GPT, Gemini or Grok extension?

No. Cline is a third-party client that connects to models from different providers. For official clients, Claude users can choose the Claude Code VS Code extension, and GPT users can choose the Codex IDE extension; each has its own setup. See Gemini in VS Code and Grok in VS Code for the editor paths supported by these Passion8 guides.

#Why does a vendor login still open after I enter my Passion8 key?

This setup uses an API key and does not require a vendor account. Check whether you selected an official-login provider or left the custom base URL disabled, and verify both Plan and Act settings. If you do not have a Passion8 key, create a token using Quickstart.

#Why does a GPT model work in Codex but fail in Cline?

Codex uses Responses, while Cline OpenAI Compatible uses Chat Completions. Whether a model ID supports both protocols depends on the current Passion8 route. Check the model and protocol in the console, then inspect the error response. For a Responses-only route, use Codex IDE setup.

#Next Steps

#Official References

• Cline OpenAI Compatible
• Cline Anthropic / Claude Code
• Cline Google Gemini
• Cline Plan and Act

• Cline Rules