Telegram founder Pavel Durov, on Monday, announced that bots will now be able to communicate with each other on the platform. The move comes at a time when agentic AI has started to make rapid inroads into the tech and commerce ecosystems. Announcing the development, Durov noted, that AI developers had been asking for this functionality which is now being delivered through this development.
This upgrade for Telegram is designed to allow bots to autonomously perform complex multi-stage tasks centred around simplifying workflows within the app’s ecosystem. In recent months, several popular apps across industries have enabled similar features as the race to AI-fueled automation gathers momentum.
Telegram is not putting bot-to-bot communications on the table for its wider ecosystem. In its announcement blog, Telegram specified particular contexts under which it will allow bots to engage with each other — creating a communication trail that humans could supervise and follow.
Breaking down Telegram’s refreshed AI feature
Telegram users interested in experimenting with the bot-to-bot communication feature will need to first activate it using the BotFeather tool. Users will need to activate this feature manually to permit their bots to send or receive messages from other bots.
Telegram said, users will be able to use the bot-to-bot communication feature under three specific environments — group chats, private chats, and Telegram business accounts.
In group chats, users will have to trigger communication trails between their bots and another one by specifically using /command@OtherBot or by sending a direct reply to a message by the other bot. If users would want their bots to access all the messages from the other bots, the “Group Privacy Mode” feature will have to be disabled and Admin Rights will have to be granted.
Just like humans, bots on Telegram will also have their own handles which will act as their prime identity on the platform.
On private chats, one bot will be able to send a direct message to another via Telegram’s standard sendMessage API and tagging the destination bot’s username. For this feature to work on Telegram’s private chats, both the bots in the equation would need the bot-to-bot feature enabled.
Telegram has claimed that unless bots are not tied to ongoing conversations, other bots will not be able to see ongoing autonomous conversations.
For users holding business accounts on Telegram, they will be allowed to set up a “primary bot” to act as a coordinator and task other secondary bots to initiate and complete tasks like bookings or customer data management.
Here’s what tricky though
While it may seem like an efficient idea to stitch AI-powered bot functions to automate everyday tasks on Telegram, the platform has flagged areas wherein the bot-to-bot communication could backfire.
Telegram said, “bot-to-bot communication can easily result in infinite interaction loops.”
In order to curb such logjams, Telegram has established certain guardrails and has asked users to adhere by the preventative instructions.
Users working with bots on the app will be able to fix rate limits on bot messages. This means, users owing the bots will be able to decide if they prefer only one reply per bot every few seconds. This will prevent a bulk of bot-to-bot messages from complicating the communication trail.
“Your bot must remain stable even if another bot intentionally responds instantly and continuously. Deduplicate repeated messages. Enforce maximum interaction depth or timeouts, both globally and per sender/receiver,” Telegram noted.
Violating these guidelines, the platform said, could lead to user restrictions and impact the overall bot ecosystem on the app.
So much of everything now about AI
In the last few months, the influx of agentic AI across sectors has become visibly more prominent.
In April itself, American fintech firm Stripe unveiled Link, a wallet service to let AI agents complete trades and transactions within themselves on behalf of their users.
Circle, the issuer of the USDC stablecoin, opened doors for agentic AI to engage with USDC. It unveiled Agent Wallets, Agent Marketplace, and Circle Skills among other features forussed on upgrading agentic AI integration with the USDC ecosystem.
Coinbase and Amazon Web Services have also joined forces to facilitate USDC micropayments via AI agents.
Charles Hoskinson, the chief of Cardano, is among several tech leaders to have predicted that AI agents could largely dominate all kinds of internet activities by 2035.
