The tool set is built around a specific loop: post the error, watch the thread, act on what a colleague replies. That needs threaded replies, sender resolution and file transfer in the same server, which is what is here. Setup is the real cost — a Cloud project, the Chat API Configuration tab, and a one-time local OAuth run before anything works. The author is explicit that this is meant for local or on-premises use inside an organisation.
A Python MCP framework for chat platforms, shipping a complete Google Chat provider. It runs against your own Google Cloud project and OAuth credentials, and the same server process can be started once per provider so an assistant reaches several platforms at once.
- get_chat_spaces_tool lists spaces; manage_space_members_tool adds and removes members; get_conversation_participants_tool returns who is in a space
- send_message_tool, reply_to_message_thread_tool, update_chat_message_tool and delete_chat_message_tool cover the message lifecycle, including replying inside an existing thread
- search_messages_tool searches history and get_my_mentions_tool surfaces the threads you were tagged in
- summarize_conversation_tool condenses a space; list_messages_with_sender_info_tool and get_message_with_sender_info_tool return messages with the sender resolved
- upload_attachment_tool, send_file_message_tool and send_file_content_tool move files into a space
- add_emoji_reaction_tool reacts to a message; batch_send_messages_tool posts several at once
- get_my_user_info_tool and get_user_info_by_id_tool resolve identities
Python 3.9 or newer with uv, plus a Google Cloud project with the Google Chat API enabled and its Configuration tab completed — the API refuses requests until that step is finished. Download credentials.json into src/providers/google_chat/ and run the local auth flow once with `python -m src.server --provider google_chat -local-auth`. Providers are declared in provider-config.yaml.