RealRelay.ai
Coding agents

Cline

Connect the Cline extension for VS Code and JetBrains to RealRelay.ai with the OpenAI Compatible provider: install it, fill in the base URL, key and model, set the model limits and check the connection.

Create an API keyBrowse the catalog
Protocol
OpenAI Chat Completions
Base URL
https://www.realrelay.ai/v1
Official docs
docs.cline.bot
Checked against the tool’s docs
2026-10

Overview

Cline is an open-source coding agent that runs inside your editor. It plans a change, edits files and runs terminal commands, and asks before each step unless you allow it to go ahead. You connect it to RealRelay.ai with its OpenAI Compatible provider, which sends requests to the Chat Completions endpoint, /v1/chat/completions.

  • An API key from the API keys page of the console.

  • The base URL with /v1 at the end. Cline adds /chat/completions to it.

  • A model ID exactly as GET /v1/models lists it, for example z-ai/glm-5.2. Pick one that supports tool calling and has at least 64K of context.

Cline cannot read a model's context window or output limit from RealRelay.ai. If you leave them empty, it assumes a 128,000-token window. Set them yourself, as shown under Configure.

Install

Install Cline from your editor's own extension marketplace. It needs no separate download.

  1. 1

    Open Extensions

    Press Ctrl+Shift+X on Windows and Linux, or Cmd+Shift+X on macOS.

  2. 2

    Install Cline

    Search for Cline and click Install. The same steps work in Cursor, Windsurf and VSCodium.

  3. 3

    Open Cline

    Click the Cline icon in the sidebar.

  1. 1

    Open the plugin marketplace

    Go to Settings → Plugins → Marketplace.

  2. 2

    Install Cline

    Search for Cline, click Install, then restart the IDE.

  3. 3

    Open Cline

    Go to View → Tool Windows → Cline.

Cline also offers its own paid account and subscription. You do not need either to use RealRelay.ai: choosing your own provider in the settings is enough.

Configure

  1. 1

    Open the settings

    Click the settings icon (⚙) at the top of the Cline panel.

  2. 2

    Choose the provider

    In the API Provider list, choose OpenAI Compatible.

  3. 3

    Fill in the connection

    Enter the base URL, paste your key in place of sk-***, and type or pick the model ID.

    Cline → Settings (⚙) → API Configuration
    1API Provider:  OpenAI Compatible2Base URL:      https://www.realrelay.ai/v13API Key:       sk-***4Model ID:      z-ai/glm-5.2
  4. 4

    Set the model limits

    Click Model Configuration to open it, then fill in the fields below. Use the numbers from the model's catalog page; the values shown are safe lower bounds, not the model's real limits.

    Cline → Settings → Model Configuration
    Supports Images:      off (on only for models that accept images)Context Window Size:  200000Max Output Tokens:    32000

  • Base URL is the one field that decides where requests go. It must end in /v1, with nothing after it.

  • Model ID keeps the part before the slash. glm-5.2 on its own is not a RealRelay.ai model ID.

  • Context Window Size tells Cline when the conversation is getting full. Max Output Tokens caps one reply.

  • Supports Images is on by default. Turn it off for text-only models, so Cline does not send them screenshots.

  • Leave Use Azure Identity Authentication off. RealRelay.ai uses the API key.

Verify

  1. 1

    Call the endpoint directly

    Run this with your key. A JSON reply with a choices array means the address, key and model work.

    Terminal
    1curl -sS "https://www.realrelay.ai/v1/chat/completions" \2  -H "Authorization: Bearer sk-***" \3  -H "content-type: application/json" \4  -d '{"model": "z-ai/glm-5.2", "max_tokens": 16, "messages": [{"role": "user", "content": "Reply with OK"}]}'
  2. 2

    Give Cline a small task

    Open a project folder, start a new task and send a request that makes Cline use a tool but changes nothing.

    Cline → New task
    List the files in this folder and say which one looks like the entry point. Do not change anything.
  3. 3

    Check the result

    Cline should ask to read the folder, then answer. That shows the model can call tools. The request also appears in the console's usage logs.

Models

To switch models, open the settings and change Model ID. Once the base URL and key are filled in, Cline loads the model list from RealRelay.ai, so you can pick from it or type an ID. Update Model Configuration each time, because the limits belong to the model.

Cline has two modes: Plan, which only reads and discusses, and Act, which makes changes. To use a stronger model for planning and a faster one for edits, turn on Use different models for Plan and Act modes in the settings, then configure each mode.

Models that work well for coding agents are listed under Choosing a model.

Troubleshooting

  • "Invalid API Key" or 401. The key is wrong or has a space at either end. Paste it again from the console.

  • 404. The base URL is missing /v1, or already ends in /chat/completions. It should end in /v1.

  • "Model Not Found". The model ID does not match GET /v1/models exactly. Check the prefix and the capital letters.

  • The model list stays empty. Cline could not read /v1/models, usually because the base URL or key is wrong. Fix those first. You can still type the model ID by hand.

  • Context errors on long tasks, or the conversation is cut short early. The context window in Model Configuration does not match the model. Set it to the model's real limit.

  • An error when you attach an image. The model is text only. Turn off Supports Images, or choose a model that accepts images.

  • Plan mode works but Act mode fails, or the other way round. Separate models for Plan and Act are turned on and one mode still has old settings. Configure both modes.

  • The model answers but never edits files or runs commands. Cline sends long prompts and relies on tool calling. Choose a model with tool calling from the catalog.