Ruslan from a bank built a Telegram bot for OpenCode: 1,298 stars and 48 contributors

On October 30, 2026, an article by Ruslan was published on Habr — a tech lead at a large bank with more than 15 years of experience in IT. He described how his open-source project opencode-telegram-bot gained 1,298 stars on GitHub and 48 contributors. According to him, before that, development had mainly been a way for him to earn money: writing code after work and on weekends did not appeal to him.

The project is a Telegram client for OpenCode, one of the CLI agents (programs that work with code from the command line). The bot runs locally and is available to only one user: their id is hardcoded in the configuration. In the chat, you can see what the agent is doing — reading and editing files, executing console commands, launching subagents (auxiliary agents within the same session). Diffs can also be viewed from a phone.

The difference from the author's previous experience is the entry point. Previously, he wrote code with AI in Cursor and Windsurf, but only in familiar languages and with review of all generated code. With CLI agents, the scheme changed: he installed Claude Code during the 2026 New Year holidays, tried OpenCode, and wanted to control the agent not only from a computer but also from a phone. That is how the bot came about.

In practice, this is an example of a narrow tool for a personal pain point. Not a universal platform, but a wrapper over a specific CLI agent with one user and one configuration. Such a project is easier to maintain: a smaller surface for errors, a clear scenario. And it is easier to attract contributors: the task is limited, and the agent's behavior is visible right in the chat.

What remains unknown: how much time went into promotion and which channels produced the main influx of stars. In the article, judging by the available fragment, this data is not present. There are also no figures on load or the number of active users — only stars and contributors.