You can put Claude Code into the telegram, and you can correct the bug by bringing a mobile phone when you go out.

By /

This Claude Code telegraph robot connects your Telegram to Claude AI, allowing you to process code items anytime, anywhere without a terminal. You can let Claude analyze, edit or interpret code, run tests, and manage files through natural dialogue, all the way from just your phone. The robot keeps a record of the conversation for each item, allowing Claude to remember the context between multiple conversations. It also has built-in security features such as user authentication and directory protection, and supports handling network hooks and scheduled tasks.
Its core advantage is mobile office efficiency: through daily use of instant messaging applications, you can get AI programming assistance at any time, turning mobile phones into development tools.

The moment Anthropic released Claude Code CLI, AI programming had evolved from “dialog box generating code” to an Agent stage that “directly took over the workflow.” But the Terminal is always a physical shackle.RichardAtCT/claude-code-telegram The emergence of this project is essentially to put top AI Agents in a layer of Telegram’s “coat” and complete the transition from CLI to Mobile-Remote-Agent The reduction of dimensions.

** Architectural logic: How does it “take over” your development environment? **

This is not a simple API call forwarding. Its underlying logic is based on Claude Code CLI。When you send instructions to Telegram Bot, it calls a smart entity in the background with file system operation permissions, Shell execution permissions, and code indexing capabilities.

  • True Contextual Awareness: Unlike the web version of Claude, which only relies on you to manually paste code snippets, this Bot resides directly in your project root directory. It has a complete File-system Access。It can grep keywords, analyze cross-file references, and locally .claude-code Maintain a complete state machine in the session.
  • Agentic Workflow: What you issue is the intent, and it performs the task. For example, if you ask it to “fix boundary holes in payment logic,” it will go through the closed loop of “read code-> run test-> find error-> modify code-> retest” on its own.

** Hardcore features: more than just “being able to chat”**

The in-depth development of the project is reflected in its adaptation to complex production environments, through simple config.json You can complete the engineer-level authority definition:

{
 "anthropic_api_key": "sk-ant-api03-xxx",
 "telegram_bot_token": "123456:ABC-DEF",
 "allowed_users": [123456789], 
 "approved_directory": "/home/user/projects/my-web-app",
 "enable_webhooks": true,
 "transcription_provider": "openai"
}
  • Rights Defense System (Security Hardening)：
    • User Whitelisting: Strictly locked allowed_usersOnly your own Telegram ID can command it.
    • Sandbox Directory: Adopted approved_directory Force the claws of AI to be locked in specific projects. Even if it is logically “autonomous”, it is also “imprisoned” in its physical path.
  • Active productivity (Webhook Ingestion): It supports integration with GitHub Webhooks. When your CI hangs up or a new Commit comes in, Bot will proactively push the message.

Why is this the “optimal solution” for mobile development?

Using SSH on your mobile phone is torture, virtual keyboard typing docker exec -it... It was a disaster. But this project changed the interactive interface toNatural Language Protocol。

You can type this command directly into the Telegram window:

/ask Check all.py files in the current directory, find database queries that do not have write exception traps, and give suggestions for refactoring.

Or more radical:

Help me write a Dockerfile, package the current project, map port 8080, and try to run it in the background. If there is an error, read the log directly and analyze the reason.

In this interactive mode, you are no longer the code farmer writing code, but the person giving instructions. Commander。

rapid deployment

If you already have a Python environment and Claude Code, deployment takes almost seconds:

#Clone Project
git clone https://github.com/RichardAtCT/claude-code-telegram.git
cd claude-code-telegram

#Install dependencies
pip install -r requirements.txt

#Configuration environment
cp config.example.json config.json
#Edit config.json and fill in your Token

#Start the robot
python main.py

5. Conclusion

RichardAtCT/claude-code-telegram Not a toy, it’s a right “Development environment is cloud-based” A hard-core practice. It breaks the cognitive bias that “you can only write code while sitting at your workstation.”

The current development paradigm is muttering: humans are responsible for defining goals and reviewing, and AI is responsible for executing paths in a controlled environment. If you want to experience the thrill of “architect at your fingertips”, it is recommended to start running on your VPS immediately.

Github：https://github.com/RichardAtCT/claude-code-telegram
Oil tubing:

Scroll to Top