How to Use Lyceum Models in Cline

Cline connects through an OpenAI-compatible provider

Cline is a coding agent for Visual Studio Code. Its OpenAI Compatible provider accepts a custom base URL, API key and model ID, so you can connect it through the extension settings.

Cline sends conversations and tool calls through an OpenAI-compatible chat interface. Lyceum reasoning models may return a reasoning trace in a separate response field, such as reasoning or reasoning_content, apart from the final answer. Check that your Cline version displays an answer and completes tool calls with the model you choose. For teams evaluating models across workloads, our analysis of the best open model APIs for agentic coding covers reasoning and output token costs.

Installing and opening the settings

Setting up Cline in Visual Studio Code requires no external build steps or configuration scripting. Developers install the extension directly through the integrated VS Code Extensions panel, configure their authentication path, and immediately access the agent sidebar.

To install the extension and access the configuration panel, execute the following steps:

1. Open Visual Studio Code and press Ctrl+Shift+X (or Cmd+Shift+X on macOS) to open the Extensions view.
2. Type Cline into the search bar, locate the official extension, and click Install.
3. Restart Visual Studio Code if the Cline icon does not appear after installation.
4. Open the Cline sidebar icon. If prompted on first launch, choose 'Use your own API key' to open provider settings.
5. If Cline was previously configured, open its settings using the gear icon.

For engineers who also use terminal-based tooling alongside VS Code, the documentation for Using Lyceum Models in opencode: Custom Provider Setup outlines an equivalent workflow for shell environments.

Filling in the endpoint and key

Once Cline settings are open, enter the provider details. If the base URL is wrong, Cline may report a connection error.

Configure the primary provider settings using the exact parameters listed below:

• API Provider: Select 'OpenAI Compatible'.
• Base URL: Enter https://api.lyceum.technology/openai/v1 as the endpoint. Do not append trailing slashes or subpaths like /chat/completions.
• API Key: Paste your API key generated from the dashboard (formatted as lk_your_api_key_here).
• Model ID: Enter the full, exact model identifier string copied from the active model catalogue, such as z-ai/glm-5.2 or moonshotai/kimi-k2.7-code.

If the connection fails, check the base URL, model ID and API key. Confirm the key has credit before running longer tasks.

Setting context and output limits by hand

The most critical step that developers skip when configuring an OpenAI-compatible provider in Cline is manual limit configuration. Under the Model Configuration section, you can customize advanced parameters like Context Window size and Max Output Tokens.

The current Lyceum models response lists model IDs but does not include context or maximum output limits. Enter both values in Cline Model Configuration. A low setting can truncate context or output; a value above the server limit can make a request fail. Use the published context windows below and check the current model card for the maximum output value. For transport differences, see what breaks when you switch models on an OpenAI-compatible API.

• z-ai/glm-5.2: 1M context window; 65,536 max output tokens
• moonshotai/kimi-k2.7-code: 256K context window; 65,536 max output tokens
• minimax/minimax-m3: 1M context window; 65,536 max output tokens
• qwen/qwen3.8-flash-next: 256K context window; 65,536 max output tokens

Enable Image Support only when the chosen model card lists image input. Keep it off for text-only models.

Checking the connection works

Before initiating an autonomous coding task inside your codebase, verify that your credentials, base URL, and model identifier are fully reachable from your local environment. Running a direct, isolated request eliminates extension-level misconfigurations and confirms endpoint availability.

Replace the example key with your own, then run this curl request. It checks authentication and a text response; it does not test Cline tool use.

curl https://api.lyceum.technology/openai/v1/chat/completions \
  -H "Authorization: Bearer lk_your_api_key_here" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "z-ai/glm-5.2",
    "messages": [{"role": "user", "content": "Respond with: OK"}],
    "reasoning_effort": "none",
    "max_tokens": 32
  }'

A successful request returns HTTP 200 and an answer in choices[0].message.content. The sample turns off reasoning so a small output limit is enough for this connection check. Then send a short task in Cline and confirm it can return an answer and use tools with your chosen model.

• HTTP 200 with answer text: The endpoint, key and model ID worked for this request.
• HTTP 401: Check the API key and Bearer header.
• Connection or model error: Check the base URL and exact model ID against the current model list.

• Cline connects to custom endpoints via the OpenAI Compatible provider setting.
• Context window limits must be set manually for each model to prevent truncation.
• Verify connection details with a terminal request before starting a workspace task.

Get an API key in the Lyceum dashboard and try it on your own repository.