Skip to content

New: Binance Agent ID support. Agents can check tokens and scam links on BNB Chain through our MCP server. Connect your agent

Docs and guides

Everything you need to get the most out of MultiScreenAI.

1. Getting started

  1. Click Get started and sign in with your Google account.
  2. Open API keys in the workspace and paste a key for any provider. The key is checked straight away and the list of models it can use loads with it.
  3. Pick models in each panel, type your message and press Enter.

One key is enough to start. A single OpenRouter key gives you hundreds of models from most labs.

2. Getting API keys

Each provider gives you a key from its own dashboard. You pay them directly for what you use. Keys are saved only in your browser.

  • OpenAI: get a key. GPT-6 Astra, GPT-6.1 Sol, GPT-6.1 Terra and the rest of the GPT family.
  • Anthropic: get a key. Claude Opus 5.5, Sonnet 5.5, Fable 5.1 and Haiku 5.5.
  • Google: get a key. Gemini 4 Argon, Gemini 3.8 Flash, the Nano Banana image models (2.1, 2, Pro, 2 Lite) and Omni Flash for video.
  • xAI: get a key. Grok 4.7, Grok 4.6, the Grok 4.20 models and Grok Imagine for images and video.
  • DeepSeek: get a key. DeepSeek V4.1 Flash and DeepSeek V4 Pro.
  • Mistral: get a key. Mistral Large, Medium, Small and Codestral.
  • Groq: get a key. Open models on LPU hardware. Often the fastest panel on screen.
  • OpenRouter: get a key. One key for hundreds of models from every major lab.
  • Perplexity: get a key. Sonar models that search the web before they answer.
  • Together AI: get a key. Hosted open models: Llama, DeepSeek, Qwen and more.
  • Fireworks: get a key. Fast inference for open models. Sync to see the full catalog.
  • Cerebras: get a key. Wafer-scale inference measured in thousands of tokens a second.
  • Moonshot: get a key. Kimi K3.1 Pro and Kimi K3, with a 1M token context.
  • Qwen: get a key. Alibaba Qwen models through Model Studio (international).
  • Meta: get a key. Muse Wonder and Muse Spark from the Meta Model API.
  • NVIDIA: get a key. Nemotron 3 models on NVIDIA NIM, plus open models from other labs.
  • Cohere: get a key. Command A+ and Command A models built for business writing, search and tools.
  • Hugging Face: get a key. One token for open models from Meta, Google, NVIDIA, Microsoft and more.
  • SambaNova: get a key. DeepSeek, Llama and gpt-oss on fast RDU chips. Free tier, no card.
  • SiliconFlow: get a key. Open models with a set of always-free ones, like Qwen3 8B and GLM-4 9B.
  • Z.ai: get a key. GLM models from Zhipu. GLM-4.7, 4.5 and 4.6V Flash are free, one request at a time.
  • Ollama Cloud: get a key. Big open models like Kimi, GLM and DeepSeek on Ollama servers. Free plan for light use.
  • Vercel AI Gateway: get a key. One key for GPT, Claude, Gemini and 400+ models, with $5 of free credit every 30 days.
  • BytePlus ModelArk: get a key. ByteDance Seedance video models: Seedance 2.5 makes clips up to 30 seconds with sound.

Press Sync all models in a panel's model menu to load every model your key can use, including models released after this page was written.

3. Multi-screen and Test modes

Multi-screen

Every panel is its own chat with its own message box. Use it to work with several AIs on different tasks at once. The bar at the bottom can still send one message to all panels.

Test

One prompt goes to every panel. Each answer is timed (first token, total time, tokens per second) and the fastest finished answer is marked. Use Blind test to hide model names and shuffle the panels, star the best answer, then press Reveal models.

Both modes support up to six panels. Use the expand button on a panel to focus it, and Add AI to add another.

4. Web search, files and images

Switch tools on in a panel's header. What is available depends on the provider's API:

  • Anthropic: web search with sources, code execution that creates downloadable files (.zip, .xlsx, .pdf, charts).
  • Google: web search with sources, code execution.
  • Perplexity: searches the web on every answer.
  • Image models such as GPT Image 2 and Nano Banana (Gemini) answer with pictures you can save.

Tools like code execution and image generation are billed by the provider to your key.

5. Attaching files

Click Attach, paste, or drop files anywhere on the page. Up to eight per message.

  • Images are resized in your browser and sent to models that can see images.
  • PDFs (up to 3 MB) are read natively by Claude, Gemini and OpenAI models. Other providers get a short note instead.
  • Text and code files (up to 400 KB) are included in the message so every model can read them.

6. Working with answers

  • Copy an answer, or download its code as a single file or a .zip of every code block.
  • Send to another panel to have a different AI check and improve an answer.
  • Merge answers asks one model to combine every panel's latest answer into one.
  • Regenerate or Stop any panel on its own.
  • Export saves the whole chat as a Markdown file. Chats are also saved in the sidebar on this device.

7. Keyboard shortcuts

  • Enter sends, Shift + Enter adds a new line.
  • Esc stops answers that are streaming, and closes menus and dialogs.

8. Troubleshooting

  • "Rejected the API key": the key is wrong, expired or has no billing set up. Replace it in API keys.
  • "Does not offer this model to your key": your account cannot use that model yet. Pick another or press Sync all models.
  • "Rate limit or quota reached": wait a moment or add credit with the provider.
  • Nothing happens after sign in: allow cookies for this site and try again.

Still stuck? Contact us.