Telegram MCP Bridge
This server acts as a read-only bridge to your personal Telegram account via MTProto, letting AI assistants browse and search your Telegram messages without sending, editing, or deleting anything.
telegram_list_chats: List Telegram cloud chats, with optionallimitand archived chat filter.telegram_get_messages: Read messages from a specific chat (newest first), with pagination vialimitandbefore_message_id, without marking them as read.telegram_search_messages: Search for text globally across all allowed chats or within a specific chat.telegram_get_message_context: Retrieve a target message along with surrounding messages (configurable before/after count) in chronological order.telegram_get_chat_info: Fetch basic metadata (name, type, member count, etc.) for a specific chat.telegram_get_image(noted in README but not in schema): Download image attachments (JPEG, PNG, GIF, WebP) as MCP image content, capped by a configurable size limit.
Key constraints:
Read-only: Cannot send, edit, delete, forward, or mark messages as read.
Access control: Can be restricted to specific chat IDs via an allowlist.
Local only: Runs locally; remote access can be facilitated through a secure tunnel.
Provides read-only access to a Telegram user account, allowing listing chats, retrieving messages, searching messages, getting message context, and getting chat information.
Click on "Install Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@Telegram MCP Bridgeshow me the last 10 messages from my Python group chat"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Telegram MCP Bridge
A local, read-only MCP bridge for accessing your own Telegram account through a user session. It uses Telethon (MTProto), not the Telegram Bot API.
Current tools
telegram_list_chatstelegram_get_messagestelegram_search_messagestelegram_get_message_contexttelegram_get_chat_infotelegram_get_image
The first version cannot send, edit, delete, forward, or mark messages as read.
Related MCP server: tdl-mcp
Requirements
Python 3.11+
Poetry
Telegram
api_idandapi_hashfrom https://my.telegram.orgAn OpenAI Platform organization with Secure MCP Tunnel access
ChatGPT developer mode / custom plugins enabled
tunnel-client.exefor Windows
Windows installation with Poetry
Clone and install the project:
cd C:\Users\YOUR_USER\Documents\Code
git clone https://github.com/boltholds/Telegram_MCP_Bridge.git
cd Telegram_MCP_Bridge
poetry installCreate the local configuration:
copy .env.example .env
notepad .envTelegram API credentials
Sign in at https://my.telegram.org.
Open API development tools.
Create an application and copy its
api_idandapi_hash.Fill in
.env:
TELEGRAM_API_ID=12345678
TELEGRAM_API_HASH=replace_me
TELEGRAM_PHONE=+79990000000
TELEGRAM_SESSION_PATH=./sessions/telegram_mcp
TELEGRAM_ALLOWED_CHAT_IDS=
TELEGRAM_MAX_MESSAGES_PER_REQUEST=100
TELEGRAM_MAX_SEARCH_RESULTS=100
TELEGRAM_MAX_MEDIA_BYTES=10485760Authorize the Telegram user session:
poetry run telegram-mcp-loginTelegram sends the code to an already authorized Telegram client. Enter the 2FA
password when requested. A successful login creates
sessions/telegram_mcp.session; this file grants access to the account and must
never be shared or committed.
Test the stdio MCP server:
poetry run telegram-mcpThe command normally stays silent and waits for MCP requests. Stop it with
Ctrl+C.
Connect ChatGPT through Secure MCP Tunnel
ChatGPT cannot invoke a local stdio process directly from a cloud conversation. Secure MCP Tunnel runs next to the bridge and opens an outbound-only HTTPS connection to OpenAI. The Telegram MCP server remains local and does not require an inbound port.
1. Create a tunnel and copy its ID
Select the same Platform organization used by the target ChatGPT account.
Create a tunnel, for example
telegram-mcp.Associate it with the target personal ChatGPT workspace (or the required Business/Enterprise workspace).
Copy the tunnel ID, which looks like
tunnel_....
The account needs Tunnels Read + Use to run and select a tunnel. Creating or
editing one additionally requires Tunnels Read + Manage.
2. Create the runtime API key
Create a runtime key at
Platform organization API keys.
The key and tunnel must belong to the same Platform organization. Do not use an
Admin API key and never put the key into the repository or .env.
Set it only in the current cmd.exe window:
set "CONTROL_PLANE_API_KEY=sk-REPLACE_ME"Verify that the variable exists without printing the secret:
if defined CONTROL_PLANE_API_KEY (echo API key is set) else (echo API key is missing)The variable disappears when the terminal closes. Set it again before future
doctor or run commands, or provide it through an appropriate local secret
manager.
3. Install tunnel-client.exe
Download the Windows tunnel client using the instructions in the Secure MCP Tunnel guide and place it somewhere local, for example:
C:\Users\YOUR_USER\Downloads\tunnel-client.exeCheck the binary:
C:\Users\YOUR_USER\Downloads\tunnel-client.exe --version
C:\Users\YOUR_USER\Downloads\tunnel-client.exe --help4. Find the Poetry Python executable
From the repository directory, run:
poetry env info --pathFor example:
C:\Users\YOUR_USER\AppData\Local\pypoetry\Cache\virtualenvs\telegram-mcp-bridge-xxxx-py3.13The Python executable is therefore:
C:/Users/YOUR_USER/AppData/Local/pypoetry/Cache/virtualenvs/telegram-mcp-bridge-xxxx-py3.13/Scripts/python.exeUse forward slashes inside the tunnel profile command. Backslashes may be
treated as escape characters and produce a broken path such as
C:UsersYOUR_USER....
Verify the module before creating the profile:
C:/Users/YOUR_USER/AppData/Local/pypoetry/Cache/virtualenvs/telegram-mcp-bridge-xxxx-py3.13/Scripts/python.exe -c "import telegram_mcp_bridge; print('OK')"5. Create the tunnel-client profile
Run from the repository directory so the child MCP process can find .env:
cd C:\Users\YOUR_USER\Documents\Code\Telegram_MCP_Bridge
C:\Users\YOUR_USER\Downloads\tunnel-client.exe init --profile telegram-mcp --tunnel-id tunnel_REPLACE_ME --mcp-command "C:/Users/YOUR_USER/AppData/Local/pypoetry/Cache/virtualenvs/telegram-mcp-bridge-xxxx-py3.13/Scripts/python.exe -m telegram_mcp_bridge.server" --open-web-uiUse one pair of double quotes around the complete --mcp-command. Unix single
quotes ('''...''') do not group arguments in Windows cmd.exe. If a broken
profile already exists, add --force:
C:\Users\YOUR_USER\Downloads\tunnel-client.exe init --force --profile telegram-mcp --tunnel-id tunnel_REPLACE_ME --mcp-command "C:/Users/YOUR_USER/AppData/Local/pypoetry/Cache/virtualenvs/telegram-mcp-bridge-xxxx-py3.13/Scripts/python.exe -m telegram_mcp_bridge.server" --open-web-uiProfiles are normally stored at:
C:\Users\YOUR_USER\AppData\Roaming\tunnel-client\telegram-mcp.yaml6. Diagnose and run the tunnel
In the same terminal containing CONTROL_PLANE_API_KEY:
cd C:\Users\YOUR_USER\Documents\Code\Telegram_MCP_Bridge
C:\Users\YOUR_USER\Downloads\tunnel-client.exe doctor --profile telegram-mcp
C:\Users\YOUR_USER\Downloads\tunnel-client.exe run --profile telegram-mcpKeep run open for connector discovery and every later MCP call. With
--open-web-ui, the local admin UI opens automatically. A healthy setup shows
the main channel as enabled, server external, and transport stdio.
7. Create the ChatGPT plugin
While tunnel-client run is active:
Open ChatGPT Settings -> Plugins -> New plugin.
Enter a name such as
Telegram MCP Bridge.Select Tunnel, not Server URL.
Select the tunnel ID created above.
No separate OAuth configuration is required by this local bridge.
Confirm the custom MCP warning and create the plugin.
Start a new chat with the plugin enabled and ask it to list Telegram chats.
When the MCP tool schema changes, update/reconnect the plugin and start a new conversation. Existing conversations may retain the older tool list.
Update and restart
cd C:\Users\YOUR_USER\Documents\Code\Telegram_MCP_Bridge
git pull
poetry installStop the running tunnel with Ctrl+C, set CONTROL_PLANE_API_KEY in the new
terminal, and run it again:
set "CONTROL_PLANE_API_KEY=sk-REPLACE_ME"
cd C:\Users\YOUR_USER\Documents\Code\Telegram_MCP_Bridge
C:\Users\YOUR_USER\Downloads\tunnel-client.exe run --profile telegram-mcpTroubleshooting
401 Unauthorized in tunnel-client logs
The control-plane key is missing, invalid, revoked, or belongs to a different
Platform organization. Stop the daemon, set a valid runtime key in the same
cmd.exe window, and restart it. Confirm the tunnel and key belong to the same
organization and the account has Tunnels Read + Use.
unknown shorthand flag: 'm' in -m
The MCP command was not enclosed in Windows double quotes. Use:
--mcp-command "C:/absolute/path/to/python.exe -m telegram_mcp_bridge.server"Executable path becomes C:Users...
Backslashes were consumed as escapes. Recreate the profile with --force and
forward slashes in --mcp-command.
Telegram session is not authorized
Stop the tunnel, authorize from the repository, then restart it:
cd C:\Users\YOUR_USER\Documents\Code\Telegram_MCP_Bridge
poetry run telegram-mcp-loginConnector creation returns Something went wrong
Keep the tunnel daemon running and inspect its local Logs page while trying
again. If no connector request reaches the daemon, verify the tunnel-to-ChatGPT
workspace association. If a request arrives and fails, run doctor and inspect
the MCP subprocess error.
Generic stdio MCP configuration
For a local MCP host that supports direct stdio processes, start the server with:
telegram-mcpExample MCP configuration:
{
"mcpServers": {
"telegram": {
"command": "/absolute/path/to/.venv/bin/telegram-mcp",
"env": {
"TELEGRAM_API_ID": "123456",
"TELEGRAM_API_HASH": "replace_me",
"TELEGRAM_SESSION_PATH": "/absolute/path/to/private/telegram_mcp"
}
}
}
}On Windows, point command to .venv\\Scripts\\telegram-mcp.exe.
Access policy
Set TELEGRAM_ALLOWED_CHAT_IDS to a comma-separated allowlist. When it is empty, all ordinary cloud chats visible to the account are accessible. For safer use, begin with one or two chat IDs returned by telegram_list_chats.
Limits are controlled with:
TELEGRAM_MAX_MESSAGES_PER_REQUEST(default: 100)TELEGRAM_MAX_SEARCH_RESULTS(default: 100)
telegram_get_image returns JPEG, PNG, GIF, and WebP attachments directly as MCP
image content. Downloads stay in memory and are capped by
TELEGRAM_MAX_MEDIA_BYTES (default: 10 MiB).
Security
The .session file grants access to the Telegram account. Never commit or share it. Keep the bridge local, use an allowlist, and review every MCP host that can invoke it.
Maintenance
Resources
Unclaimed servers have limited discoverability.
Looking for Admin?
If you are the server author, to access and configure the admin panel.
Latest Blog Posts
- Your AI Chatbot Just Exposed Your CEO's Salary to an InternBy Om-Shree-0709 on .Agent IdentityMCP SecurityOAuth Delegation
- Why MCP Servers Need Execution Sandboxing (And Why Your Current Stack Isn't Enough)By Om-Shree-0709 on .Agentic AiPrompt InjectionWebAssembly
MCP directory API
We provide all the information about MCP servers via our MCP API.
curl -X GET 'https://glama.ai/api/mcp/v1/servers/boltholds/Telegram_MCP_Bridge'
If you have feedback or need assistance with the MCP directory API, please join our Discord server