Chat right here on the web with streaming replies and your full history, or connect a Telegram bot in about a minute and talk to the same agent from your phone. It keeps one memory across every channel.

Flip on Council mode and two different AI models each draft a reply independently — then the two are merged into one more considered response. It uses more credits per message, so you turn it on only when the answer really matters.

Give your agent a recurring job — a morning briefing, a daily summary — and it runs on its own schedule, delivering the result here or straight to Telegram. Set it once and forget it.

Ask your agent for a picture and it generates one on the spot — no separate tool, no setup. The result appears right in your chat and is saved to your agent’s Files for later.

Send a picture of a receipt, a whiteboard, a homework problem, or a document and your agent reads the text and understands what’s in it — right in web chat or over Telegram. No scanning app, no copy-typing.

One agent, a lot of range — all in its own private, secure sandbox.
Your agent builds a lasting memory of you and your business across every conversation and channel.
Upload menus, price lists, FAQs or notes and your agent always has them at hand — or add a searchable document library it quotes from with citations, page numbers and all.
Put several agents on a panel with a moderator and hand them a topic. They discuss it live and conclude with one worked-out answer.
Every agent runs isolated in its own container — its data, memory and secrets never mix with anyone else’s.
Give a colleague or client access to one agent — chat-only or read-along — while your own conversations stay private. Revoke anytime, instantly.
Done with an agent for now? Archive it to secure cold storage and stop paying for hosting — then restore it later with its memory, chats and documents intact.
Export any conversation, or download everything your agent knows — memory, chats and knowledge — whenever you want. No lock-in.
Built-in guardrails stop prompt-injection and credential-fishing before they ever reach your agent.
Your plan keeps agents always-on; usage runs on prepaid credits — $1 for 10 credits, and new accounts start with 100 free.
Kick the tires
For regular use
For power users
Answer questions, help customers, research, write, remember context, generate images, and read images you send it — using your uploaded knowledge and its own memory. Advanced mode adds tools like web search and files inside its sandbox.
Yes. Attach a photo in web chat or send one over Telegram — a receipt, a whiteboard, a homework problem, an ID, a screenshot — and your agent reads the text (OCR) and understands what’s in the image, then answers about it. No separate scanning app.
Chat with it yourself on the web, or connect a Telegram bot and share it. You approve each person who pairs with the bot.
A switch for higher-stakes questions. Instead of one model answering, two different AI models each draft a reply independently and the two are merged into a single, more considered response — a built-in second opinion. It uses more credits per message, so you turn it on only when the answer really matters.
Yes, two ways. Team seats (Business) share all your agents with teammates. Or share a single agent with anyone by email — chat-only or with read-along access to its files and documents. They chat in their own private conversation, you can revoke access instantly, and their usage bills your credits with a cap you control.
Idle agents go to sleep automatically, and you can archive one entirely: it moves to secure cold storage, stops using hosting, and no longer counts against your plan’s agent limit. Restore it anytime — memory, conversations and documents come back exactly as they were.
Each agent runs in its own isolated container with its own encrypted secrets and private storage. Nothing is shared between agents or customers.
Each plan includes a set number of always-on agents for a flat monthly fee. What your agents actually use is metered as prepaid credits (1 credit ≈ 20,000 tokens); top up anytime at $1 for 10 credits. New accounts start with 100 free credits.
Agents run on DeepSeek by default, with automatic context management so long conversations keep working — you can switch models in each agent’s settings.