ToRun Chat - One Interface for Every AI Model, Mode, and Modality

Pick the job. We pick the model.

Most AI chat apps lock you into one provider's model and one way of working. ToRun Chat flips that around. You tell us what you're trying to do - write, research, look at an image, narrate a script, run a snippet of code - and ToRun routes your message to the model that's genuinely best at it, across every major provider.

No more memorizing model names. No more "is this the one that can see images?" You choose a mode. We handle the rest - and show you exactly what each message costs before it adds up.

Start chatting free · See what's included in each plan

One window, sixteen modes

Chat isn't a single feature - it's a workspace with a mode for every kind of work. Switch among them from the same conversation:

Chat - everyday multi-turn conversation across every provider.
Code - write, explain, and run code, with a sandbox to execute it inline.
Image - generate artwork and visuals from a prompt without leaving the thread.
Video - turn a script into video clips.
Audio & Voice - narrate text, transcribe speech, or hold a real-time spoken conversation.
Deep Research - multi-step, cited investigation (its own showcase).
Canvas - a longer-form writing and editing surface.
Council - put a question to several models at once and compare their answers side by side.
Translator - move text and speech between languages.
Persona - chat as or with an AI persona that has its own voice and memory.
Workflow & Agent - kick off automations and autonomous tasks.
Embed (RAG) - ground answers in your own uploaded knowledge.

You don't manage a fleet of tools. You pick the job in front of you, and the right engine shows up.

Any model, one conversation - switch mid-chat

OpenAI, Claude, Gemini, Mistral, Perplexity, DeepSeek and more all live behind the same chat window. Start a thread on one model and switch to another in the very next message - your full history comes along, so there's no re-explaining and no copy-paste between tabs.

Because routing is capability-first, you never have to know which model supports what. Ask for vision, and ToRun quietly selects a model that can actually see. Ask for deep research, and it picks one built for multi-step reasoning. Ask for reasoning, long context, or a specific language, and the router filters to models that deliver it - then chooses by price for your tier. The right engine, every time, without you managing a fleet of subscriptions.

Why it matters: you get the strongest answer for each task instead of forcing one model to do everything - and if a provider has an outage or raises prices, you're never stranded. Zero lock-in, by design.

Talk, type, show, and listen

Chat is fully multi-modal, so a single thread can move fluidly between the way you want to communicate:

See - drop in images or video and ask questions about them. Vision-capable models read the content and answer in context.
Speak and hear - real-time voice in, natural voice out, over a live low-latency connection. Or attach a recording and have the model work from the transcript.
Create - generate images right inside the conversation, then iterate on them in the same thread.
Read sources - pull in live web results, with citations, when you need current information.

It's one continuous conversation that adapts to the moment, instead of five separate tools you have to stitch together.

Run code without leaving the chat

Ask for a script and Code mode doesn't just print it - it can execute it in a sandbox and show you the output inline. Generate a chart, parse a file, test a regex, or check an algorithm, and see the real result in the thread instead of copying code into a separate editor. The sandbox is isolated and ephemeral, and every run that uses a model is billed like any other call - transparently.

Research you can actually cite

Switch into Deep Research mode and a simple question becomes a sourced answer. ToRun runs a multi-step investigation - searching the web, checking your knowledge base, cross-referencing findings - then hands back a synthesized response where every claim links to its source, quote and URL included.

Findings from multiple sources are de-duplicated and ranked by confidence, so you're reading verified material, not a confident guess. It's the difference between an answer you hope is right and one you can paste into a report, a brief, or a paper.

Ask the room, not just one model

Some questions deserve more than one opinion. Council mode sends your prompt to several models at once and lays their responses side by side, so you can compare reasoning, catch where they disagree, and pick the answer you trust - or have the platform adjudicate a consensus. It's a second (and third) opinion, on demand.

Memory that stays fresh, and a cost you can see

ToRun Chat remembers what matters about you and keeps it honest over time. Tell it your birthdate once and it always knows your current age - recomputed, never a stale number you have to correct. Facts that drift are flagged with "as of a few months ago - may be stale" instead of being repeated as gospel. You stay in control with a memory editor where you decide what's remembered.

And every single message carries its true price. Each turn produces one itemized billing record - the model used, the capabilities billed, the exact amount - so AI spend is never a mystery at the end of the month. No hidden meters, no surprise invoices, just transparent cost you watch in real time.

A good conversation shouldn't be trapped in one window. Share a chat with a public link so others can read it, let them remix it into their own thread, and branch a conversation to explore a different direction without losing the original. Your history is searchable and yours to revisit.

What you can do with ToRun Chat

Switch models mid-chat bi-arrow-left-right - move between OpenAI, Claude, Gemini, Mistral, Perplexity and more inside one thread; your full history follows, so you never re-explain or copy-paste between tabs.
Pick a mode, not a model bi-ui-checks-grid - choose from sixteen modes - Chat, Code, Image, Video, Voice, Deep Research, Council, Translator and more - and capability-first routing selects the best engine for that exact job.
See, speak, and listen bi-camera-video - send images and video, talk in real time with natural voice replies, and generate images right inside the conversation.
Run code inline bi-code-slash - execute snippets in a sandbox without leaving the chat and watch the output appear in the thread.
Compare several models at once bi-people - Council mode asks multiple models the same question and lays the answers side by side so you can choose the one you trust.
Research you can cite bi-quote - Deep Research returns multi-step answers where every claim links to its source - verified, de-duplicated, ranked by confidence.
See the cost of every message bi-receipt - each turn produces one itemized charge - model, capabilities, exact amount - so spend is transparent in real time, with no vendor lock-in.