Telegram Launches Serverless for Bots, Targets AI Coding Agents

Telegram has introduced Serverless, a hosting service for bot and Mini App backends that removes the need to run your own server. Code deploys with a single command, and the platform is explicitly aimed at developers working with AI agents, not just people writing code by hand.

How Serverless Works

Developers write plain JavaScript modules split into three parts: handlers that respond to events (messages, button clicks, inline queries), a shared code library, and a single file defining the database schema. The tgcloud command-line tool syncs the local project with the cloud and provides commands such as push, migrate, run, status, and diff.

When an update reaches the bot, Telegram routes it to the handler file matching the event type, for example handlers/message.js or handlers/callback_query.js. The handler has access to three SDK elements: db for database queries, api covering the full Bot API, and fetch for calls to external interfaces.

Why It Targets AI Agents

Telegram's documentation explicitly addresses developers who work with AI assistants. Every new Serverless project includes an AGENTS.md file and a docs/tgcloud-sdk.md reference file, which agentic tools read automatically to learn the project's conventions, such as importing by name without relative paths, or the two-step process for deploying database changes through push and migrate.

Prefer to build with an AI assistant - or is the only coder on your team an AI? You can still ship a bot. - from Telegram Serverless documentation

The new-project scaffolding even includes ready-made instructions for launching a specific tool, such as opencode, Claude Code, or Cursor, as an entry point for an agent expected to read the AGENTS.md file. The tgcloud run command lets developers test a handler locally without deploying, shortening the feedback loop for an agent writing code through trial and error.

Limitations and Architecture

The platform is deliberately stripped down. There is no way to install npm packages or import external dependencies, no file system access, and the only support for media files is reusing an existing Telegram file_id. The database runs on SQLite with foreign keys disabled, schema migrations are kept separate from code deployment to avoid accidental data loss, and the system detects conflicts when multiple developers work on the same project.

The documentation does not list pricing or usage limits, aside from the hard 32 MB cap on handler responses. Telegram mainly emphasizes that developers no longer need to rent, update, and monitor their own machine.

Market Context

Serverless for Telegram bots puts the messenger in direct competition with general-purpose serverless platforms such as AWS Lambda, Cloudflare Workers, and Deno Deploy, at least in the segment of simple bot integrations. Analysts note that Telegram's strategy is to capture projects at the moment an AI agent generates the first code and deploys it with a single command, before the developer has a chance to consider alternative infrastructure.

For Polish bot developers and integrators, this opens a new, cheaper path for running simple Telegram automations, especially in projects driven largely by coding agents, without the need to set up separate hosting or manage a production server.