Telegram MCP Server

Telegram MCP Server

Enables AI assistants to interact with users via Telegram to request decisions, approvals, or specific input through text and clickable buttons. This facilitates a human-in-the-loop workflow where the AI can pause for feedback or send status notifications during long-running tasks.

Category
Visit Server

README

Telegram MCP Server

License: MIT Python 3.10+ Docker MCP Telegram Bot API

A Model Context Protocol (MCP) server that enables AI assistants (like Kilo Code) to ask you questions via Telegram and wait for your responses. This creates a "human-in-the-loop" workflow where the AI can request decisions, approvals, or specific input during long-running tasks.

Features

  • šŸ¤– Interactive AI Workflow: AI can pause and ask you questions via Telegram
  • šŸ“± Button Support: Present multiple choice options as clickable buttons
  • ā±ļø Long Polling: Waits up to 2 minutes for your response
  • šŸ”’ Secure: Uses environment variables for credentials
  • šŸŽÆ Simple Integration: Works with any MCP-compatible AI assistant
  • 🐳 Docker Support: Run natively or in a container

Prerequisites

  • Python 3.10+ (for native installation)
  • OR Docker and Docker Compose (for containerized installation)
  • A Telegram account
  • A Telegram Bot Token (from @BotFather)

Installation Methods

You can run this MCP server in two ways:

  1. Native Python Installation - Run directly on your system
  2. Docker Installation - Run in a container (recommended for production)

Native Installation

1. Create a Telegram Bot

  1. Open Telegram and search for @BotFather
  2. Send /newbot and follow the prompts
  3. Name your bot (e.g., "MyDevBot")
  4. Copy the HTTP API Token provided
  5. Important: Send /start to your new bot so it can message you

2. Find Your Telegram User ID (Optional)

  1. Search for @userinfobot on Telegram
  2. Send it any message
  3. Copy your User ID

3. Clone and Setup

# Clone the repository
git clone https://github.com/yourusername/telegram-mcp-server.git
cd telegram-mcp-server

# Create virtual environment
python3 -m venv venv

# Activate virtual environment
source venv/bin/activate  # On Linux/Mac
# OR
venv\Scripts\activate  # On Windows

# Install dependencies
pip install -r requirements.txt

4. Configuration

Create a .env file from the sample:

cp .env.sample .env

Edit .env and add your credentials:

TELEGRAM_BOT_TOKEN=your_bot_token_here
TELEGRAM_USER_ID=your_user_id_here  # Optional - bot will auto-detect if not set

āš ļø Security Note: The .env file is gitignored to protect your credentials.

5. Test the Server Locally

# Activate virtual environment
source venv/bin/activate

# Run the server
python telegram_server.py

You should see:

╭──────────────────────────────────────────────────────────────────────────────╮
│                         ▄▀▀ ā–„ā–€ā–ˆ ā–ˆā–€ā–€ ā–€ā–ˆā–€ ā–ˆā–€ā–„ā–€ā–ˆ ā–ˆā–€ā–€ ā–ˆā–€ā–ˆ                        │
│                         ā–ˆā–€  ā–ˆā–€ā–ˆ ā–„ā–„ā–ˆ  ā–ˆ  ā–ˆ ā–€ ā–ˆ ā–ˆā–„ā–„ ā–ˆā–€ā–€                        │
│                                FastMCP 2.14.2                                │
│                    šŸ–„  Server:      Telegram Human Loop                       │
╰──────────────────────────────────────────────────────────────────────────────╯

Press Ctrl+C to stop the test.

Docker Installation

1. Configure Environment

cp .env.sample .env
# Edit .env with your credentials

2. Build and Run

# Build the Docker image
docker-compose build

# Start the container in detached mode
docker-compose up -d

3. View Logs

docker-compose logs -f telegram-mcp

4. Stop the Server

docker-compose down

Docker Commands Reference

Building:

# Build the image
docker-compose build

# Build without cache (force rebuild)
docker-compose build --no-cache

Running:

# Start in detached mode (background)
docker-compose up -d

# Start in foreground (see logs directly)
docker-compose up

# Start and rebuild if needed
docker-compose up -d --build

Monitoring:

# View logs
docker-compose logs -f telegram-mcp

# View last 100 lines of logs
docker-compose logs --tail=100 telegram-mcp

# Check container status
docker-compose ps

# Execute commands inside the container
docker-compose exec telegram-mcp python --version

Stopping and Cleaning:

# Stop the container
docker-compose stop

# Stop and remove containers
docker-compose down

# Stop, remove containers, and remove volumes
docker-compose down -v

# Remove all (containers, networks, images)
docker-compose down --rmi all

Restart:

# Restart the container
docker-compose restart

# Rebuild and restart from scratch
docker-compose down
docker-compose build --no-cache
docker-compose up -d

Development:

# Open a shell in the running container
docker-compose exec telegram-mcp /bin/bash

# Test the server locally (without Docker)
source venv/bin/activate && python telegram_server.py

Connecting to MCP Clients

For Kilo Code (or similar MCP clients)

Method 1: Using the Startup Script (Recommended)

  1. Open your MCP settings file:

    • Location: ~/.config/Code/User/globalStorage/rooveterinaryinc.roo-cline/settings/mcpSettings.json
    • Or use your client's UI: Settings → MCP Servers → Edit Configuration
  2. Add this configuration (replace /absolute/path/to/telegram-mcp-server with your actual installation path):

{
  "mcpServers": {
    "telegram": {
      "command": "bash",
      "args": ["/absolute/path/to/telegram-mcp-server/run.sh"]
    }
  }
}

Example: If you cloned to /home/user/projects/telegram-mcp-server, use:

{
  "mcpServers": {
    "telegram": {
      "command": "bash",
      "args": ["/home/user/projects/telegram-mcp-server/run.sh"]
    }
  }
}
  1. Restart your MCP client or reload the MCP servers

Note: An example configuration is provided in mcp-config.example.json for reference.

Method 2: Direct Python Execution

{
  "mcpServers": {
    "telegram": {
      "command": "/absolute/path/to/telegram-mcp-server/venv/bin/python",
      "args": ["/absolute/path/to/telegram-mcp-server/telegram_server.py"]
    }
  }
}

Method 3: Using Docker

Option A: Docker Compose Exec

{
  "mcpServers": {
    "telegram": {
      "command": "docker-compose",
      "args": [
        "-f",
        "/absolute/path/to/telegram-mcp-server/docker-compose.yml",
        "exec",
        "-T",
        "telegram-mcp",
        "python",
        "telegram_server.py"
      ],
      "cwd": "/absolute/path/to/telegram-mcp-server"
    }
  }
}

Option B: Docker Run

{
  "mcpServers": {
    "telegram": {
      "command": "docker",
      "args": [
        "run",
        "--rm",
        "-i",
        "--env-file",
        "/absolute/path/to/telegram-mcp-server/.env",
        "telegram-mcp-server",
        "python",
        "telegram_server.py"
      ]
    }
  }
}

Available Tools

Once connected, your AI assistant will have access to these tools:

  1. ask_human(question, options, wait, timeout_seconds, allow_custom) - Send questions to Telegram

    • wait=True (default): Blocks until you respond or timeout
    • wait=False: Sends question and returns immediately (non-blocking mode)
    • options: Optional list of button choices
    • allow_custom=True (default): Adds "Custom answer" button when options provided
    • Works in both blocking and non-blocking modes
  2. get_telegram_response(mark_as_read) - Retrieve your latest Telegram message

    • Use after ask_human(wait=False) to get your response
    • mark_as_read=True (default): Won't retrieve the same message twice
  3. send_telegram_notification(message) - Send one-way status updates

    • For progress reports, completion notifications, or status updates
    • Doesn't expect a response
    • Perfect for keeping you informed during long-running tasks
  4. list_telegram_messages(limit) - View recent conversation history

    • Shows last 5-20 messages for context
    • Useful for checking if you've already replied

Usage Examples

Example 1: Simple Question

You say to your AI:

"I'm going to grab coffee. If you need to know which database to use, ask me via Telegram."

AI calls:

ask_human(question="Should I use PostgreSQL or MongoDB for this project?")

What happens:

  1. Your phone buzzes with a Telegram message
  2. You reply: "PostgreSQL"
  3. AI receives "PostgreSQL" and continues

Example 2: Multiple Choice with Buttons

AI calls:

ask_human(
    question="How should I structure the authentication?",
    options=["JWT", "Session Cookies", "OAuth2", "Skip for now"]
)

What happens:

  1. You receive a Telegram message with 4 clickable buttons
  2. You tap "JWT"
  3. AI receives "JWT" instantly

Example 3: Approval Workflow

You say to your AI:

"Refactor the entire codebase, but ask me before making any breaking changes."

AI calls:

ask_human(
    question="I want to rename `getUserData()` to `fetchUser()`. This will break 23 files. Proceed?",
    options=["Yes, proceed", "No, skip this", "Show me the files first"]
)

Example 4: Non-Blocking Mode for Complex Questions

AI asks without waiting:

ask_human(
    question="Please review this 500-line refactor and provide detailed feedback",
    wait=False
)

What happens:

  1. You receive the question on Telegram
  2. You take your time to review (no timeout)
  3. When ready, you tell the AI: "I've answered on Telegram"
  4. AI retrieves your answer: get_telegram_response()

Example 5: Progress Notifications

AI sends status updates:

send_telegram_notification("šŸš€ Starting database migration...")
send_telegram_notification("āœ… Step 1/5 complete: Schema created")
send_telegram_notification("šŸŽ‰ Migration complete! All 1,247 records migrated successfully.")

How It Works

  1. AI calls ask_human() with a question and optional button choices
  2. Server sends Telegram message to your configured chat
  3. Long polling loop checks Telegram every 2 seconds for your response
  4. You respond via text or button click
  5. Server returns your answer to the AI
  6. AI continues with your input

Timeout Behavior

  • Default timeout: 120 seconds (2 minutes)
  • Configurable via the timeout_seconds parameter in ask_human()
  • If you don't respond in time, the AI receives: "Timeout: User did not respond in time..."

Non-Blocking Mode

For complex questions that require extended thinking time, use non-blocking mode to avoid timeouts:

  1. AI asks without waiting:

    ask_human("Please review this architecture and provide feedback", wait=False)
    
  2. You receive the question and can take as long as needed to think and respond

  3. When ready, tell the AI you've replied:

    "I've answered on Telegram"

  4. AI retrieves your answer:

    answer = get_telegram_response()
    

Custom Answers with Buttons

When you provide button options, the AI automatically adds a "āœļø Custom answer" button (unless allow_custom=False). This lets you:

  • Click a button for quick selection
  • OR type your own custom response

Example:

ask_human(
    "Which database?",
    options=["PostgreSQL", "MongoDB", "MySQL"],
    wait=False
)

You'll see 4 buttons:

  • PostgreSQL
  • MongoDB
  • MySQL
  • āœļø Custom answer (type below)

Best Practices

Use Non-Blocking Mode (wait=False) when:

  • Questions require code review or deep analysis
  • You might be away from your device
  • The decision requires research or consultation
  • You need more than 2 minutes to respond

Use Blocking Mode (wait=True) when:

  • Questions are simple yes/no decisions
  • You're actively working with the AI
  • Quick responses are expected
  • Using button options for multiple choice

Use Notifications (send_telegram_notification) for:

  • Progress updates during long-running tasks
  • Task completion notifications
  • Error or warning alerts
  • Status updates that don't require a response
  • Keeping yourself informed while away from the computer

Project Structure

telegram-mcp-server/
ā”œā”€ā”€ telegram_server.py         # Main MCP server code
ā”œā”€ā”€ .env                       # Your credentials (gitignored)
ā”œā”€ā”€ .env.sample               # Template for environment variables
ā”œā”€ā”€ .gitignore                # Protects secrets
ā”œā”€ā”€ requirements.txt          # Python dependencies
ā”œā”€ā”€ run.sh                    # Startup script (executable)
ā”œā”€ā”€ mcp-config.example.json   # Example MCP configuration (customize for your setup)
ā”œā”€ā”€ Dockerfile                # Docker image definition
ā”œā”€ā”€ docker-compose.yml        # Docker Compose configuration
ā”œā”€ā”€ .dockerignore             # Docker build exclusions
└── README.md                 # This file

Troubleshooting

"Command 'python' not found"

Solution: Use python3 instead, or install the symlink:

sudo apt install python-is-python3

The run.sh script already uses the correct Python from the virtual environment.

"Error: Could not find a Telegram Chat ID"

Solution: Make sure you've sent /start to your bot at least once.

"Permission denied: ./run.sh"

Solution: Make the script executable:

chmod +x run.sh

"Timeout: User did not respond in time"

Solutions:

  • Respond faster (within 2 minutes)
  • Increase timeout in the ask_human() call
  • Use non-blocking mode (wait=False) for complex questions

MCP Server Not Showing in Client

Solutions:

  1. Check the MCP settings file path is correct
  2. Verify the absolute path in your configuration
  3. Restart your MCP client completely
  4. Check client logs for connection errors

Docker Container Won't Start

Check logs:

docker-compose logs telegram-mcp

Environment Variables Not Loading

Ensure .env file exists and is properly formatted:

cat .env

Permission Issues with Docker

If you encounter permission issues, ensure the .env file is readable:

chmod 644 .env

Advanced Configuration

Changing the Default Timeout

You can customize the timeout per question:

# Short timeout for quick questions
ask_human("Proceed?", options=["Yes", "No"], timeout_seconds=30)

# Longer timeout for thoughtful questions
ask_human("Which approach?", timeout_seconds=300)  # 5 minutes

Restricting to Specific User

The server uses TELEGRAM_USER_ID from .env. Only messages from this user ID will be accepted.

Using Multiple Bots

Create separate directories with different .env files and register each as a different MCP server:

{
  "mcpServers": {
    "telegram-work": {
      "command": "bash",
      "args": ["/path/to/work-bot/run.sh"]
    },
    "telegram-personal": {
      "command": "bash",
      "args": ["/path/to/personal-bot/run.sh"]
    }
  }
}

Security Best Practices

  1. āœ… Never commit .env - Already in .gitignore
  2. āœ… Use TELEGRAM_USER_ID - Prevents unauthorized users from controlling your AI
  3. āœ… Regenerate tokens if accidentally exposed
  4. āœ… Use private Telegram bots - Don't share your bot with others
  5. āœ… Use Docker secrets for sensitive data in production
  6. āœ… Regularly update the base Python image for security patches

Production Deployment

For production deployment:

  1. Remove development volume mounts from docker-compose.yml
  2. Use environment variables or secrets management instead of .env file
  3. Consider using a process manager or orchestration tool (Kubernetes, Docker Swarm)
  4. Set up proper logging and monitoring
  5. Configure restart policies appropriately

Example production docker-compose.yml:

version: '3.8'

services:
  telegram-mcp:
    build:
      context: .
      dockerfile: Dockerfile
    container_name: telegram-mcp-server
    environment:
      - TELEGRAM_BOT_TOKEN=${TELEGRAM_BOT_TOKEN}
      - TELEGRAM_USER_ID=${TELEGRAM_USER_ID}
      - PYTHONUNBUFFERED=1
    restart: always
    logging:
      driver: "json-file"
      options:
        max-size: "10m"
        max-file: "3"

Dependencies

  • fastmcp (2.14.2+) - MCP server framework
  • httpx (0.28.1+) - Async HTTP client for Telegram API
  • python-dotenv (1.2.1+) - Environment variable management

See requirements.txt for full list.

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

License

MIT License - Use freely!

Support

For issues with:

Changelog

v1.0.0 (2026-01-02)

  • Initial release
  • Support for blocking and non-blocking question modes
  • Button support for multiple choice questions
  • Progress notification system
  • Docker support
  • Comprehensive documentation

Recommended Servers

playwright-mcp

playwright-mcp

A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.

Official
Featured
TypeScript
Magic Component Platform (MCP)

Magic Component Platform (MCP)

An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.

Official
Featured
Local
TypeScript
Audiense Insights MCP Server

Audiense Insights MCP Server

Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.

Official
Featured
Local
TypeScript
VeyraX MCP

VeyraX MCP

Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.

Official
Featured
Local
Kagi MCP Server

Kagi MCP Server

An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.

Official
Featured
Python
graphlit-mcp-server

graphlit-mcp-server

The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.

Official
Featured
TypeScript
E2B

E2B

Using MCP to run code via e2b.

Official
Featured
Neon Database

Neon Database

MCP server for interacting with Neon Management API and databases

Official
Featured
Exa Search

Exa Search

A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.

Official
Featured
Qdrant Server

Qdrant Server

This repository is an example of how to create a MCP server for Qdrant, a vector search engine.

Official
Featured