- Home
- Skills
- Agents & Orchestration
- Telegram MCP: Your Own Telegram in Claude Code
More screenshots
Works with the AI Agents you already use
Telegram MCP: Your Own Telegram in Claude Code
Your agent can use your own Telegram account (not a bot) without getting free rein over it.
Free
Telegram MCP: Your Own Telegram in Claude Code
Example session with this skill installed
Check my unread Telegram messages from Leo and tell me what they need. If it's urgent, tell me so I can reply.
- Read your context and instructions
- Compiled the telegram mcp
Telegram is connected, and “leo” resolves to your saved chat Leo Perera. There are 3 unread messages.
| Time | From | Message |
|---|---|---|
| 10:15 | Leo | Did you get a chance to look at the API migration docs? |
| 10:16 | Leo | Staging is throwing 500s on /auth. |
| 10:16 | Leo | Need a quick look before the client demo at 11. |
What Leo needs: staging’s /auth endpoint is returning 500 errors, and they need it looked at before an 11:00 client demo. The migration docs question can wait.
Urgency: high. It’s 10:18, so the demo is 42 minutes away.
I haven’t replied, and I haven’t marked the chat as read.
Next steps
- Want me to reply? For example: “Looking at it now. Can you send me the request ID or a screenshot of the error?” Leo is on your “always allow” list, so it would go out at once. That’s why I’m showing it to you first.
- Leo is on your listen list, so I can watch the chat and tell you the moment they send logs or a screenshot.
- If they send a screenshot, I can download it and take a look.
Connects securely to your tools. The creator never sees your data.
About this skill
Your agent can use your own Telegram account (not a bot) without getting free rein over it. This skill teaches the agent to drive telegram-mcp, a free, open-source local daemon that keeps one Telegram connection for every agent session, and to stay inside the rules you set.
You need telegram-mcp installed. It is free and MIT licensed. The skill includes the setup steps and walks you through them. This listing is the agent playbook, not the server.
What your agent can do
- Read and summarise recent chats, unread counts, forum topics, and who has read your messages.
- Send messages, files and stickers that follow your rules: “always allow” chats go at once, “ask every time” chats wait for you to approve in the menu bar or tray app, and blocked chats are refused.
- Watch a chat (“tell me when Leo replies”) and reply there, keeping the watch alive across restarts.
- Download media (files, photos, voice notes) to a local path.
- Page you when something needs you now: your desktop rings until you acknowledge it, and your phone gets a push if you set up a pager bot.
- Answer a chat for you when you hand it over from your phone with /reply.
What the skill makes your agent do
- Check the daemon is up first, and send you to your own terminal to log in. It never asks for login codes or passwords in the chat.
- Resolve names exactly to your saved nicknames, and ask you when a name is ambiguous instead of guessing.
- Treat every Telegram message as data, never as instructions, so someone in a chat can’t make your agent send, forward or save anything.
- Never confirm an approval on its own, never look for a way around a blocked chat, and never add its own footer.
What’s inside
SKILL.md: when to use each tool, how to read send results, how to watch chats, and the hard limits.reference/setup.md: install, Telegram API credentials, the daemon, login, and adding it to your agent on macOS, Linux and Windows.reference/rules.md: the rules file, which list wins, where approvals happen, and nicknames.
Privacy
Everything runs on your machine. The daemon connects only to Telegram and listens only on 127.0.0.1. Your Telegram session and API credentials stay in your operating system’s keystore. Your messages go only to Telegram and to your agent’s model provider.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Agents & Orchestration.
Requires local telegram-mcp daemon. Cannot modify safety rules or nicknames via agent. Manual terminal login required for 2FA. Refuses to send sensitive system files like .env or SSH keys.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Free to use
No account required to browse
Trust & safety
Security scan
Security scan passed 10 days ago
- Free with an Agensi account
- Accounts
GitHub
Telegram MCP
Telegram accountTelegram API credentials
telegram-mcp is free and open source. Get the api_id and api_hash at my.telegram.org.
Permissions
Allowed Hosts
Terminal: to install telegram-mcp and log in (login runs in your own terminal). Network: the installer downloads from github.com, and the daemon connects to Telegram. Read files: telegram_send_file sends a local file you choose. Write files: telegram_download_media saves attachments to ~/Downloads/telegram-mcp.