telegram-mcp
Bidirectional bridge between Telegram and MCP-compatible agents, allowing agents to poll for new messages and send replies through Telegram, without any built-in language model.
README
telegram-mcp
Bidirectional Telegram ⇆ agent bridge. No LLM in our stack.
Your agent (Claude Desktop, Cursor, a custom script) is the brain. Our platform is the pipe: we take messages you send on Telegram, store them, and hand them to your agent when it polls. When your agent has something to say back, it pushes through us to your Telegram.
- Bring your own agent. Any MCP client works.
- No inference in the middle. We don't call an LLM. No vendor lock-in.
- Bidirectional.
get_new_messagespulls inbound,send_telegram_messagepushes outbound. - Platform slash-commands.
/help,/clear,/recent,/window,/usageall work even if your agent is offline.
Hosted at telegram.regiq.in or self-host with the Docker Compose below.
Use it (hosted)
-
Visit https://telegram.regiq.in, sign in with Google.
-
DM @BotFather →
/newbot. Paste the token in your dashboard. -
On the dashboard: Generate MCP API key, copy it.
-
Add to Claude Desktop's
claude_desktop_config.json(or Cursor / any MCP client):{ "mcpServers": { "telegram": { "url": "https://telegram.regiq.in/api/mcp", "headers": { "Authorization": "Bearer YOUR_KEY_HERE" } } } } -
Send
/startto your bot on Telegram.
Now your agent can call get_new_messages to see what you sent it, and send_telegram_message to reply.
Tools
| Tool | Direction | Notes |
|---|---|---|
get_new_messages() |
user → agent | Returns inbound messages since the last pull. Read-once. |
send_telegram_message(text) |
agent → user | Delivers text to the user's Telegram bot. |
list_recent_messages(max=20) |
read | Last N stored messages, both directions. No cursor advance. |
clear_history() |
action | Wipes stored history for the user. |
Slash commands (platform-handled)
Handled by the platform directly — your agent isn't involved. So /help works even when nothing is polling.
| Command | Effect |
|---|---|
/start |
Registers this Telegram chat with your account. |
/help |
Lists commands. |
/recent [N] |
Shows the last N stored messages. |
/clear |
Wipes history. |
/window N |
Sets the history window (default 20). |
/usage |
Stored message count. |
Self-host
You'll need:
- Docker + Compose
- A Google OAuth client — Cloud Console. Scopes:
openid,email,profile. Redirect:<PUBLIC_BASE_URL>/api/auth/callback/google - A public HTTPS domain for Telegram webhooks (Cloudflare Tunnel or
ngrok http 3012locally)
No LLM key required. Your agent brings its own model.
git clone https://github.com/globalion/telegram-mcp
cd telegram-mcp
cp .env.example .env
# fill NEXTAUTH_SECRET, GOOGLE_*, PUBLIC_BASE_URL
docker compose up -d
# open http://localhost:3012
Architecture
Telegram app Your MCP agent
▲ │ ▲ │
│ │ text poll │ │ push
│ ▼ │ ▼
┌──────────────────────────────────────────────────┐
│ Our platform (Hetzner) │
│ - /api/telegram/webhook/<per-user-secret> │
│ · slash-commands answered here │
│ · plain text stored, no reply │
│ - /api/mcp Bearer-auth JSON-RPC │
│ · get_new_messages, send_telegram_message │
│ · list_recent_messages, clear_history │
│ - Postgres: User, ApiKey, TelegramLink, Message │
└──────────────────────────────────────────────────┘
Per-user webhook URL segment is a random secret so a leaked bot token can't be pointed at someone else's account. Google sign-in is identity + ban anchor only — per-message identity comes from the MCP API key + chatId.
License
MIT. See LICENSE.
Recommended Servers
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.
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.
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.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
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.
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.
Neon Database
MCP server for interacting with Neon Management API and databases
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.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.
E2B
Using MCP to run code via e2b.