Telegram Bot API Docs MCP

FastMCP server that gives AI clients access to the current Telegram Bot API documentation. The server scrapes the official docs on startup, validates the method/type graph, and serves the result through MCP tools and JSON resources.

Run

Run the published package over stdio, which is the usual transport for local AI client bindings:

uvx telegram-bot-api-docs-mcp

Or run as an HTTP MCP server:

uvx telegram-bot-api-docs-mcp --transport http --host 127.0.0.1 --port 8080

Startup requires internet access because the server fetches https://core.telegram.org/bots/api every time it starts.

For local development from a checkout, use uv run python main.py with the same arguments.

Codex

This repo includes a project-scoped Codex config at .codex/config.toml:

[mcp_servers.telegram_bot_api_docs]
command = "uv"
args = ["run", "python", "main.py"]
cwd = ".."
enabled = true
startup_timeout_sec = 60
tool_timeout_sec = 120

Start a new Codex session from this trusted repo and run /mcp to confirm that telegram_bot_api_docs is connected.

Tools

• get_bot_api_overview: version, release date, changelog URL, and counts.
• list_methods / list_types: paginated method/type discovery with optional search.
• get_method / get_type: full docs for a method or type.
• get_field: docs for a specific method parameter or type field.
• search_bot_api: search names, descriptions, fields, return types, and relationships.
• get_related_types: subtype and usage graph for a type.
• dump_bot_api_spec: raw full spec or a single section.

Resources

• telegram-bot-api://overview
• telegram-bot-api://methods
• telegram-bot-api://types
• telegram-bot-api://spec

Scraper

scrape.py contains the parser and schema validation used by the MCP server. Running it directly still writes api.json and api.min.json, but those files are ignored because the MCP server serves freshly scraped data.