mcp-telegram

mcp-telegram

MCP server that connects AI assistants to your real Telegram account via User API (MTProto). Features default-deny ACL with per-chat permissions, message search, file sending, forwarding, media downloads, and rate limiting.

Category
Visit Server

README

mcp-telegram

MCP Server Go License: MIT

An MCP server that connects AI assistants like Claude to your real Telegram account via the User API (MTProto). Not a bot — Claude reads and sends messages as you.

Built with gotd/td and the official MCP Go SDK.

Telegram API Terms of Service: This project uses the Telegram User API. You must obtain your own api_id and api_hash from my.telegram.org and comply with the Telegram API Terms of Service. Misuse of the User API (spam, bulk messaging, scraping) may result in your account being banned. You are solely responsible for how you use this tool.

Contents


Features

Tool What it does Required permission
tg_me Returns current account info
tg_dialogs Lists dialogs visible to the ACL whitelist
tg_history Fetches message history with pagination, date filtering, and media download read
tg_search Searches messages in a chat by text query, with optional sender filter read
tg_send Sends a text message or file, with optional reply-to send
tg_forward Forwards messages from one chat to another read + send
tg_draft Saves a draft message (does not send) draft
tg_mark_read Marks a chat as read mark_read

Additional capabilities:

  • File and photo sending
  • Forward messages between chats
  • Reply to specific messages
  • Download photos and documents from message history
  • Filter history by date range (since / until)
  • Typed peer references (user:ID, chat:ID, channel:ID) to prevent ID collisions
  • Lazy peer resolution — avoids FLOOD_WAIT errors at startup
  • Global rate limiting at the RPC level

What you can do with it

Once connected, you can ask your AI assistant things like:

Catch up on messages

  • "Check my unread Telegram messages and give me a summary"
  • "What did @alice write in the last 24 hours?"
  • "Show me messages from the Dev Team chat since Monday"

Reply and communicate

  • "Draft a response to the last message from @bob — don't send it yet"
  • "Send 'sounds good, let's meet at 3pm' to @alice"
  • "Reply to message 1234 in the project chat with my feedback"

Manage your inbox

  • "Mark all read in the news channel"
  • "Which of my whitelisted chats have unread messages?"
  • "Download the photos from today's messages in the design chat"

Research and analyze

  • "Find all messages mentioning the deployment in the last week"
  • "Summarize the discussion in the team chat from yesterday"
  • "What files were shared in the project channel this month?"

How it compares to chaindead/telegram-mcp

mcp-telegram chaindead/telegram-mcp
Access control Default-deny ACL whitelist with granular per-chat permissions Full access to all chats
Peer addressing Typed references (user:ID, chat:ID, channel:ID) Numeric IDs only (collision-prone)
Configuration YAML config with environment variable expansion CLI flags
Startup safety Lazy peer resolution (no bulk API calls) Eager resolution (FLOOD_WAIT risk)
Rate limiting Built-in token bucket middleware None
File support Send files, photos; download media from history Text only
Reply support Yes No
Date filtering Yes No

Quick start

Prerequisites

  • Go 1.26+
  • A Telegram account
  • API credentials from my.telegram.org (api_id and api_hash)

Install

Homebrew (macOS / Linux):

brew install Prgebish/tap/mcp-telegram

NPX (no install needed):

npx @prgebish/mcp-telegram serve --config config.yaml

Pre-built binaries (macOS / Linux / Windows):

Download from GitHub Releases.

Go install:

go install github.com/Prgebish/mcp-telegram/cmd/mcp-telegram@latest

From source:

git clone https://github.com/Prgebish/mcp-telegram.git
cd mcp-telegram
go build ./cmd/mcp-telegram

This produces mcp-telegram (or mcp-telegram.exe on Windows) in the current directory.

Authenticate

Run the auth command once to create a session file. You will be prompted for your phone number, the login code, and (if enabled) your 2FA password.

macOS / Linux:

export TG_APP_ID=12345
export TG_API_HASH="your_api_hash"

mcp-telegram auth --config config.yaml

Windows (PowerShell):

$env:TG_APP_ID = "12345"
$env:TG_API_HASH = "your_api_hash"

mcp-telegram.exe auth --config config.yaml

Windows (cmd):

set TG_APP_ID=12345
set TG_API_HASH=your_api_hash

mcp-telegram.exe auth --config config.yaml

Configure

Create a config.yaml:

telegram:
  app_id: ${TG_APP_ID}
  api_hash: ${TG_API_HASH}
  session_path: ~/.config/mcp-telegram/session.json

acl:
  chats:
    - match: "@username"
      permissions: [read, draft, mark_read]
    - match: "user:123456789"
      permissions: [read, send]
    - match: "channel:2225853048"
      permissions: [read, mark_read]

limits:
  max_messages_per_request: 50
  max_dialogs_per_request: 100
  rate:
    requests_per_second: 2.0
    burst: 3

logging:
  level: info

Environment variables in ${...} syntax are expanded at load time.

Client configuration

The server communicates over stdio — your MCP client starts and manages the process.

Claude Code (CLI — add via command):

claude mcp add telegram -- /path/to/mcp-telegram serve --config /path/to/config.yaml

Claude Desktop / Claude Code (~/.claude.json or claude_desktop_config.json):

{
  "mcpServers": {
    "telegram": {
      "command": "/path/to/mcp-telegram",
      "args": ["serve", "--config", "/path/to/config.yaml"],
      "env": {
        "TG_APP_ID": "12345",
        "TG_API_HASH": "your_api_hash"
      }
    }
  }
}

Cursor (Settings > MCP Servers > Add):

{
  "telegram": {
    "command": "/path/to/mcp-telegram",
    "args": ["serve", "--config", "/path/to/config.yaml"],
    "env": {
      "TG_APP_ID": "12345",
      "TG_API_HASH": "your_api_hash"
    }
  }
}

Configuration

ACL

The ACL is default-deny. Only chats explicitly listed in acl.chats are accessible, and only with the permissions you specify.

Supported match patterns:

Pattern Example Description
@username @johndoe Match by Telegram username (case-insensitive)
+phone +79001234567 Match by phone number
user:ID user:123456789 Match a user by numeric ID
chat:ID chat:987654321 Match a group chat by numeric ID
channel:ID channel:2225853048 Match a channel or supergroup by numeric ID

Permission types: read, send, draft, mark_read.

If the same peer matches multiple rules (e.g. via @username and user:ID), permissions are merged — they never shadow each other.

Rate limiting

The limits.rate section configures a global token bucket that wraps all Telegram RPC calls:

  • requests_per_second — sustained rate (default: 2.0)
  • burst — maximum burst size (default: 3)

Media download

media:
  download: [photo, document, video, voice, audio]
  directory: ~/telegram-media
  allowed_upload_dirs:
    - ~/Documents
    - ~/Downloads

When configured, tg_history will automatically download media files to the specified directory. The download_to parameter can override the path, but only to media.directory or its subdirectories.

allowed_upload_dirs restricts which directories tg_send can read files from. File sending is disabled unless this is configured.


Security

  • Default-deny ACL — no chat is accessible unless explicitly whitelisted
  • Filesystem boundarytg_send can only read files from allowed_upload_dirs; download_to is restricted to subdirectories of media.directory
  • Session file permissions — enforced to 0600 (owner-only read/write)
  • No secret logging — API hashes, session tokens, and auth keys are never written to logs
  • No access hash exposure — internal Telegram access hashes are stripped from all tool output
  • Rate limiting — prevents accidental API abuse
  • Local timezone — date filters use your system timezone, not UTC

License

MIT


If you find this project useful, please give it a star — it helps others discover it.

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
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
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
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