Overview

ToRun is built around one idea: you should think about the task, not the model. Instead of asking "which model is best for this?", you pick a mode — a product surface that matches what you want to do — and ToRun's routing pipeline handles model selection automatically.

What is a Mode?

A mode defines the kind of work you're doing. Current modes include:

• Chat — open-ended conversation, Q&A, brainstorming
• Code — software generation, explanation, debugging, and review
• Image — text-to-image generation
• Research — web-grounded answers with source citations
• Canvas — long-form document drafting and editing (rich editor, AI inline assist)
• Video / Music / Audio — generative media creation

Each mode declares the set of capabilities it requires — for example, Code mode needs structured output and code execution; Research mode needs web search and retrieval.

How Capability Routing Works

Once you choose a mode, the pipeline runs:

1. Identify required capabilities — each mode has a fixed capability set (text, vision, tools, code execution, image generation, search, reasoning, and more).
2. Filter the model catalog — only models that support all required capabilities are considered.
3. Rank by price and quality — the best available match for your plan and wallet balance is selected automatically.
4. Fallback if needed — if a provider is down or over capacity, the next-best model takes over without interrupting your session.

When to Switch Modes

Tips

• You can switch modes mid-session. Your conversation history carries over where it makes sense.
• If a mode is unavailable on your current plan or wallet balance, ToRun tells you exactly why — no silent failures.
• Advanced users can bring their own API keys (BYOK) to route through their own provider accounts, still within the same mode and routing system.

Modes are the entry point. The routing is the engine. You stay focused on the work.