Documentation
Talking to your agent
Talking to your agent works just like messaging any Telegram contact — no new app, no new interface to learn.
What it understands
| You send | The agent does |
|---|---|
| Text | Replies normally, with context from the conversation and from memory. |
| Voice message | Transcribes it automatically and replies as if it were text (requires the transcription service from Step 5). |
| Photo | Sees the image: describes it, analyzes it, extracts text from it, answers questions about it. |
| Document (PDF, spreadsheet…) | Saves the file and can read and work with its contents. |
Multiple messages in a row
Sent 3 messages while the agent was thinking? No problem: it bundles them together and replies once, taking all of them into account — just like a person would.
While it’s working
- The “typing…” indicator means it’s processing. Longer tasks (research, document analysis) can take a few minutes.
- Messages sent while it’s busy go into a queue and are answered next.
- Want to cancel? Send /stop — it cancels the current task and clears the queue.
Tip: just talk naturally
No special commands or formatting needed. “remind me to pay the bill tomorrow at 9am”, “summarize this PDF”, “what did I tell you about Tuesday’s meeting?” — plain language is exactly what it expects.Files have an expiration date
Photos, documents, and generated images are not kept forever: uploaded files are cleaned up periodically, and images created by the agent are deleted after 30 days. Want to keep something permanently? Just ask in chat: “save this to the library” — library files are never deleted and the agent can find them in any future conversation.