Every deployed Somodus agent supports two-way Telegram chat, including voice: send a voice message and the agent transcribes it, does the work, and can reply with spoken audio. By voice you can ask questions, run the agent's actions, and adjust or pause its schedule — from your phone, with the regular Telegram app.
Your agent answers with its actual tools — it can search the web, look up X (Twitter) posts and trends, pull YouTube transcripts, and read back its own past reports. "What did you send me on Monday?" gets a real answer, not a canned reply.
Whatever actions your agent was built with — sending a report, posting to a channel, updating a spreadsheet — you can trigger from chat or voice. For actions that publish or send something, the agent confirms with you before it acts.
"Run at 7am instead of 8." "Pause until Monday." The agent updates its own schedule or pauses and resumes itself — no dashboard visit needed.
Tell it "always answer in Korean," "focus on enterprise news," or "remember that my launch is in March" — the agent saves your instructions, language, location, and notes, and applies them from then on.
Send a voice message in Telegram. The agent transcribes it with speech-to-text, reasons about what you asked using OpenAI's frontier models and its built-in tools, does the work, and replies in text — and, for voice messages, it can also reply with spoken audio, so the whole exchange works hands-free. Scheduled work is unaffected: your agent keeps running its regular jobs on its own schedule, and the chat is how you steer it between runs.
Most no-code chatbot builders treat Telegram as a text channel. Voice is where they stop: as of July 2026, two-way voice-message handling — transcribing an incoming Telegram voice note and replying with spoken audio — is not something mainstream bot builders document as a built-in capability; typical guidance is to wire up your own speech-to-text integration. With Somodus it is built in for every deployed agent, with nothing to configure.
The deeper difference is what's behind the chat. A chatbot builder gives you a conversation flow you design. A Somodus agent is an autonomous worker that already runs your task on a schedule — the Telegram chat (text or voice) is a remote control for a worker that exists anyway, with real tools: web search, your integrations, its own run history, and its own schedule.
Yes, within clear limits: by voice you can ask questions, run the agent's actions, adjust or pause its schedule, and update its standing instructions. Scheduled runs stay on their schedule — voice control steers the agent rather than replacing the schedule.
No. It works in the regular Telegram app on any phone or desktop — you connect your agent to your Telegram account once, then just send messages or voice notes.
Text always; for voice messages it can also reply with spoken audio, so a voice question gets a spoken answer you can listen to hands-free.
Yes. Voice transcription and spoken replies are part of the AI usage included in every plan — chatting with your agent consumes plan credits like any other AI work, with no separate speech API to set up.
Agents reply and report in English, Korean, Japanese, Spanish, and Simplified and Traditional Chinese — speak to your agent in your language and it answers in kind.
What you can automate with a Claude AI agent — the full catalog of agent use cases, and Playable content your agent creates — mini-games and interactive pages, delivered on schedule.