Features
Six tools. One environment.
All models. One chat.
Claude, GPT, Gemini, Llama, Mistral, Grok. Plus the open-source frontier. Switch mid-conversation, context stays.
- New models appear on launch day.
- No new tab, no new account.
- The new model sees the full history.
Conversations that branch
Try ten directions in parallel, the original stays. Every branch inherits the context.
- Nested as deeply as you like.
- Full-text searchable.
- Delete a branch, the original stays.
Voice
Dictate prompts instead of typing them. Answers are read aloud.
Presets and prompts
Reuse your own system prompts and prompt library with a click.
Memory
What you tell the AI about yourself, it remembers. Across all chats.
Agents that really work
Define model, tools and rules once. From then on the AI delivers on its own.
- Web search and MCP tools out of the box.
- Your own tools via function calling.
- Share or export as a template.
Images from a prompt
From sentence to finished image, without leaving the chat.
- Images with DALL-E, Imagen, Stable Diffusion.
- Variations directly from history.
Files that get understood
PDFs, Excel, screenshots, source code. Drag, drop, ask. No RAG pipeline needed.
- Up to 200 MB per file, multiple in parallel.
- Tables read as data, not as image.
- Code with indentation and syntax.
Full-text search
Search all chats for words, snippets or filenames.
Artifacts
HTML, React components and SVG rendered live in the chat. See instead of copy.
Share via link
Share chats via link, with or without login required.
Code, cleanly formatted
Python, JavaScript, Shell and dozens more. In clean, syntax-highlighted code blocks right in the chat.
- Syntax highlighting for dozens of languages.
- Copy with one click.
- Long snippets stay readable and scrollable.
One API. Same credit.
OpenAI-compatible across every model. Same account, same billing as in the chat.
Drop-in compatible
Existing SDKs keep working. Just swap base URL and key.
Streaming and tools
SSE, function calling, structured outputs. Unified format across every provider.
Multiple keys, one account
One key per project with its own limit. Shared credit with your chat.
Cost down to the cent
You see after every prompt what it cost. No flat rate, no estimate.
-
01
Real cost, not estimated
You see the exact amount per prompt. Not "approximately", but exactly.
-
02
Full breakdown
Per request: input tokens, output tokens, model, price. Also for images and audio.
-
03
Limits that work
Daily or monthly cap. Warning at 75 percent, stop at 100.