Don't pick a model. Describe the task.

The Agent reads your prompt and picks the right AI model for the job, chat, image, video, music, code, search. It routes to top-tier models: Opus 5, GPT-5.5 Pro, Veo 3.1, Sora 2 Pro, Nano Banana Pro. Included on every plan.

Try the Agent free ✦ See pricing

1. You describe the task

"Generate a 9:16 vertical clip of a fox running through snow with cinematic camera move." Or "Write a one-liner that explains AGI to a kid." Or "Translate this PDF to Russian and summarise the key points."

2. SmartRouter classifies it

Reads the prompt, attached files, conversation history. Picks the right category (chat / image / video / audio / code / search) and the right model at your tier. Takes ~1 second, costs 2 tokens.

3. The model runs, you see the result

For sync tools (chat, image, music) the result appears in the chat. For async (video) you get a job_id, the agent reminds you when it's ready. Both surface in the same UI.

One agent, the top model for every task

💬 Text, code & search

Each prompt is routed to a top-tier model.

  • Chat: Claude Opus 5, GPT-5.5 Pro, Gemini 3 Pro
  • Code: GPT-5 Codex, Claude Sonnet 4.6, DeepSeek R1
  • Search: Perplexity Sonar Pro, Grok

Included on every plan and the Free tier. No tier to choose, the Agent always routes to the best available model.

Test the Agent with €0.99

1,000 tokens covers hundreds of agent picks plus the actual model runs. No model picker, no separate subscriptions per tool.

Start now ✦
FAQ

Agent questions

How does the Agent pick a model?
SmartRouter reads your prompt and any attached files, classifies the task (chat / image / video / audio / code / search), then routes it to the top model for that category: Opus 5 for chat, Veo 3.1 and Sora 2 Pro for video, Nano Banana Pro for image, Seedance 2.0 for fast video. The pick takes about 1 second and costs 2 tokens.
Which models does the Agent use?
Every category routes to a top-tier model: Opus 5, GPT-5.5 Pro or Gemini 3 Pro for chat, Veo 3.1 / Sora 2 Pro / Kling v3.0 Pro for video, Nano Banana Pro, GPT-5 Image or Recraft v3 for image, ElevenLabs Music for audio. Same prompt, the best available quality.
Does the Agent cost extra?
No, the Agent itself is included on every plan (Free, Starter, Beginner, Fullhouse, Complete, VIP). The router's decision costs 2 tokens per pick, then the chosen model bills at its normal rate. There is no separate Agent subscription.
Can I still pick a specific model?
Yes. Open the model picker on any chat to override the Agent's pick. Useful when you want a specific model (e.g. DeepSeek R1 for reasoning) or need to compare outputs. The picker shows the per-call cost so you can choose.
Where can I use the Agent?
Web app at avots.ai/chat, Telegram bot @AvotsAIbot, and through MCP from Claude.ai web, Claude Desktop, Claude Code CLI, Cursor, Cline.
Start creating