RealRelay.ai
Coding agents

AionUi

Connect AionUi's built-in agent to RealRelay.ai through its New API platform: start from the console or add the platform by hand, then check the connection and pick models.

Create an API keyBrowse the catalog
Protocol
OpenAI Chat Completions · Anthropic Messages
Base URL
https://www.realrelay.ai/v1
Official docs
github.com
Checked against the tool’s docs
2026-10
Experimental. Not yet run end to end against RealRelay.ai. The steps follow the tool’s own documentation; if one fails, contact support with the error.

Overview

AionUi is a free, open-source desktop app where AI agents work on your computer: they read and write files, run tasks and use tools, with your approval. It ships with a built-in agent, and that agent is what you connect to RealRelay.ai. AionUi's New API platform sends each model's requests to /v1/chat/completions by default, or to /v1/messages for models you set to the Anthropic protocol.

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

  • The base URL with /v1. The New API platform also accepts the site root and fixes the path itself.

  • A model ID exactly as GET /v1/models lists it, for example z-ai/glm-5.2.

AionUi can also drive Claude Code, Codex and other CLI agents it finds on your computer. Those agents keep their own settings and do not use the models you add here; set them up with their own guides or with CC Switch. This setup has not yet been run end to end against RealRelay.ai.

One-click setup

The console has an aionui:// link that opens AionUi's model settings. Install AionUi first (see Install below) so your browser knows how to open it.

  1. 1

    Open the key's menu

    On the API Keys page, open the ⋯ menu in the Actions column of the key you want to use, then choose Chat → AionUI. Allow the browser to open the app.

  2. 2

    Complete the Add Model dialog

    AionUi opens Settings → Models with the Add Model dialog. In AionUi 2.2 the console's link does not fill in the address and key, because the two use different field names. Fill in the dialog as shown under Configure, then click Confirm.

The link saves you only the navigation, so if your console has no Chat → AionUI item, the manual steps below are just as quick.

Install

Install AionUi with Homebrew on macOS, or download the installer for your system from GitHub Releases. It needs macOS 10.14, Windows 10, or a current Linux such as Ubuntu 18.04 or later, with at least 4 GB of memory.

macOS (Homebrew)
brew install aionui

Or download the .dmg for Apple Silicon or Intel, open it and drag AionUi into Applications.

Download the .exe installer and follow the prompts.

Install the .deb with your package manager, or make the .AppImage executable and run it.

Configure

  1. 1

    Open the model settings

    Click the settings icon in the left sidebar, open Models, and click Add Model.

  2. 2

    Choose New API and fill in the fields

    Set Model Platform to New API, then enter the endpoint, your key and the first model. Leave Full URL off so AionUi adds the request path. Replace sk-*** with your key.

    Settings → Models → Add Model
    1Model Platform: New API2API Endpoint:   https://www.realrelay.ai/v13Full URL:       off4API Key:        sk-***5Model Name:     z-ai/glm-5.2
  3. 3

    Set how the model is called

    For each model, AionUi stores a Request Protocol. Use OpenAI for everything except Claude models, and set OpenAI request API to Chat Completions; RealRelay.ai serves only a few OpenAI models on the Responses API.

    Settings → Models → model settings
    Model:              z-ai/glm-5.2Request Protocol:   OpenAIOpenAI request API: Chat Completions
  4. 4

    Save

    Click Confirm. RealRelay.ai now appears in the model list with the model you added.

  • New API is the platform AionUi provides for New API gateways such as RealRelay.ai. It lets each model use its own protocol and accepts the base URL with or without /v1.

  • The Custom (OpenAI Compatible) platform also works if you only use Chat Completions models. With it, the endpoint must end in /v1.

    Settings → Models → Add Model
    1Model Platform: Custom (OpenAI Compatible)2API Endpoint:   https://www.realrelay.ai/v13API Key:        sk-***4Model Name:     z-ai/glm-5.2

  • The API Key field takes several keys, separated by commas or line breaks. AionUi rotates them and skips a failing key for a short time.

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

    Run the health check

    In Settings → Models, run Health Check on the model. A latency figure means AionUi reached RealRelay.ai.

  3. 3

    Send a message

    Start a new chat with the built-in agent, pick the RealRelay.ai model in the model selector at the bottom left of the input box, and send a short message. The request then appears in the console's usage logs.

Models

Add more models to the RealRelay.ai platform from its entry in Settings → Models. You can pick from the list AionUi loads from RealRelay.ai, or type any model ID. Switch models in a conversation with the selector at the bottom left of the input box.

  • The built-in agent calls tools, so choose models with tool calling. AionUi may hide models without it from the conversation picker.

  • Claude models can use the Anthropic protocol, which RealRelay.ai serves natively. AionUi selects it on its own for IDs that start with anthropic:

    Settings → Models → model settings
    Model:            anthropic/claude-sonnet-4-6Request Protocol: Anthropic

  • Do not set models to the Gemini protocol. RealRelay.ai has no Gemini text model.

For help picking a model, see Choosing a model.

Troubleshooting

  • The Add Model dialog opens empty after using the console link. Expected in AionUi 2.2; fill in the fields by hand as above.

  • The model list is empty or fails to load. The key is wrong or disabled, or the endpoint has a typo. Check the key on the API keys page and the address, then try again.

  • 404. Full URL is on, or, with the Custom platform, the endpoint is missing /v1. Turn Full URL off and use the endpoint shown above.

  • An error about the model or the endpoint for an OpenAI model. AionUi sent it to the Responses API. Set OpenAI request API to Chat Completions, unless the model is one RealRelay.ai serves on Responses.

  • A configured model does not show up in the conversation. The model lacks tool calling, or the settings have not reloaded. Choose a model with tool calling, save again and restart AionUi.

  • Claude Code or Codex inside AionUi ignores RealRelay.ai. Those agents use their own configuration. Set them up separately, for example with CC Switch.