About Telegram Archive Mcp
Telegram Archive Mcp is an MCP server for Telegram-Archive instances, intended for LLMs that need to search and browse archived chats. It supports chat, topic, date, and message-context queries, and can return attached photos.
Key features
- 01Search messages across or within chats
- 02Browse chat history with pagination
- 03Read messages by date and timezone
- 04View forum topics and pinned messages
- 05Retrieve message photos and file details
Use cases
- 01Find a specific archived message
- 02Browse chat history for a given day
- 03Review nearby messages and attached photos
MCP support
It provides MCP tools over stdio and Streamable HTTP and connects to a Telegram-Archive instance. Tools are read-only except for refreshing statistics, which can also be hidden with READ_ONLY=1.
Configuration
Local server config is the same for every client. Copy it into your client’s MCP config file.
npx -y telegram-archive-mcpTELEGRAM_ARCHIVE_URLTelegram Archive instance URL (e.g. http://localhost:8000)TELEGRAM_ARCHIVE_USERTelegram Archive login usernameTELEGRAM_ARCHIVE_PASSSecretTelegram Archive login passwordREAD_ONLYSet to true to hide refresh_stats, the one tool that asks the viewer to do workLocal servers need a runtime on your machine (e.g. Node.js 18+). Config file locations differ by client; see your client’s docs.
Tools
Local servers expose tools only once running on your machine, and the maintainer has not provided a tool list.
Details
From the project README
telegram-archive-mcp
An MCP server for any Telegram-Archive instance. LLMs use it to search messages, browse chats and read archived Telegram history.
Features
- Speaks MCP 2026-07-28 and every earlier revision down to 2024-11-05 on stdio and Streamable HTTP. Clients that still use the
initializehandshake work unchanged. - Search one chat or every chat at once, read a chat page by page, one calendar day in your timezone, or the conversation around one message.
- Forum topics, pinned messages, folders and per-chat statistics.
- The photo attached to a message, returned as an image the model can look at and scaled to 1568 pixels when larger (
get_media). - Every tool is read-only except
refresh_stats;READ_ONLY=1hides that one too. - Signs in to Telegram-Archive through
/api/loginwithTELEGRAM_ARCHIVE_USERandTELEGRAM_ARCHIVE_PASS. - HTTP listens on loopback by default, checks
Origin, caps request bodies, and requiresMCP_AUTH_TOKENon any other address. - Ships as an npm package, a Docker image and multi-arch Go binaries.
Install
Every client below runs the server over stdio. Set the URL, user and password of your Telegram-Archive viewer.
Claude Code
claude mcp add --env TELEGRAM_ARCHIVE_URL=http://localhost:8000 \
--env TELEGRAM_ARCHIVE_USER=admin --env TELEGRAM_ARCHIVE_PASS=your-viewer-password \
--transport stdio telegram-archive -- npx -y telegram-archive-mcp
Claude Desktop (claude_desktop_config.json) and Cursor (~/.cursor/mcp.json)
{
"mcpServers": {
"telegram-archive": {
"command": "npx",
"args": ["-y", "telegram-archive-mcp"],
"env": {
"TELEGRAM_ARCHIVE_URL": "http://localhost:8000",
"TELEGRAM_ARCHIVE_USER": "admin",
"TELEGRAM_ARCHIVE_PASS": "your-viewer-password"
}
}
}
}
VS Code (.vscode/mcp.json, asks for the password once)
{
"inputs": [
{ "type": "promptString", "id": "tg-pass", "description": "Telegram-Archive viewer password", "password": true }
],
"servers": {
"telegram-archive": {
"type": "stdio",
"command": "npx",
"args": ["-y", "telegram-archive-mcp"],
"env": {
"TELEGRAM_ARCHIVE_URL": "http://localhost:8000",
"TELEGRAM_ARCHIVE_USER": "admin",
"TELEGRAM_ARCHIVE_PASS": "${input:tg-pass}"
}
}
}
}
Codex (~/.codex/config.toml)
[mcp_servers.telegram-archive]
command = "npx"
args = ["-y", "telegram-archive-mcp"]
env = { TELEGRAM_ARCHIVE_URL = "http://localhost:8000", TELEGRAM_ARCHIVE_USER = "admin", TELEGRAM_ARCHIVE_PASS = "your-viewer-password" }
Docker, for any client above: use docker as the command with these arguments, and a release tag from the releases page, never latest:
docker run -i --rm -e TELEGRAM_ARCHIVE_URL -e TELEGRAM_ARCHIVE_USER -e TELEGRAM_ARCHIVE_PASS drumsergio/telegram-archive-mcp:vX.Y.Z
The npm package downloads the release binary on first start and checks it against the checksums published with the package, so it also works with npm 12, which skips install scripts. A long-running HTTP server (Docker Compose) and local builds are in Getting started.
Tools
| Tool | What it does |
|---|---|
list_chats | Lists chats, most recent first; filter by search (name), archived or folder_id. Each chat's ref is the chat_id the other tools take. |
list_folders | Lists your Telegram folders and their ids. |
search_all_chats | Searches message text in every chat at once. |
search_messages | Searches one chat, optionally one forum topic. |
get_messages | Reads a chat newest first, by offset or keyset cursor, optionally one forum topic. |
get_message_context | Returns one message with the messages before and after it. |
get_messages_by_date | Returns every message of one calendar day in a timezone. |
get_pinned_messages | Returns a chat's pinned messages. |
get_topics | Lists a forum group's topics; their ids feed topic_id. |
get_chat_stats | Message and media counts and size of one chat. |
get_media | The photo attached to a message, as an image, or the file's details. |
refresh_stats | Makes the viewer recalculate archive-wide statistics. Hidden with READ_ONLY=1. |
Every tool returns structured content with an output schema, alongside the same JSON as text. Arguments and defaults are on Usage.
Documentation
The full documentation is at geiserx.github.io/telegram-archive-mcp.
- Getting started: npm, Docker Compose, local build
- Configuration: environment variables and client configuration
- Usage: the 4 resources and 12 tools, paging, reading one day at a time
- Development: testing, contributing, credits
- Related projects: Telegram Archive, other MCP servers, where this one is listed
License
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.