Until now this was blocked at the platform level — bots couldn't see messages from other bots at all, in any mode. The FAQ even says so explicitly: they decided to block it to prevent two bots from going into an infinite reply loop.
Now bot-to-bot communication works in groups and via business mode. Enabled through @BotFather.
You can now build bot chains inside Telegram without an external orchestrator.
A coordinator bot receives a task, calls an analytics bot, which passes the result to a generator bot. The whole pipeline — in one chat.
It won't replace n8n, Make, or Langchain. But a native feature always comes in handy.
Business mode in the Bot API arrived in March 2024 (Bot API 7.2), and even then bots got access to private chats of business accounts. Bot-to-bot is the next step: if a bot could act on behalf of a business, now bots can coordinate with each other.
How did they solve the infinite loop problem (Loop Prevention)?
Telegram put the responsibility on developers. The platform requires implementing safeguards: message deduplication, hard timeouts, and rate limits. If two bots go haywire and start spamming each other, Telegram will apply platform-level restrictions and throttle performance.
In short, you can now assemble a whole agency or editorial team of bots in one chat:
🤖 Agent storms. In one chat you can set up a copywriter bot, a designer bot, and an editor bot. You throw in a topic, they argue, generate content, fact-check each other, and deliver a finished post right before your eyes.
🔹 Microservices right inside the messenger. No more building one bloated Frankenstein bot. Split the functionality: one bot handles billing (has access to the cash register), another parses data, a third works with an LLM. If the neural network goes down, payments and support keep running.
🔹 Mutual AI oversight. The main bot talks to the client in a group or Business mode, while a hidden supervisor bot "listens" to the responses. If the first bot starts hallucinating, the second one instantly takes over the conversation and calls in a human manager.
🔹 Database isolation. The public bot collects requests from clients but has no access to your internal CRM or warehouse. It tags a closed internal bot (`/check_stock@MyStoreBot`), which checks the data and safely returns the result.
In short, you can now build entire autonomous AI teams.