Chat agent that keeps memories and notes in Google Docs

What it is

This n8n workflow gives you a chat agent that remembers you and keeps your notes. You talk to it in n8n’s chat panel. Before it answers, it reads two Google Docs, one of memories and one of notes. It adds to them when you share something to keep. Each answer is also copied to Telegram. Thirteen nodes do the work, beside eight sticky notes.

Who it suits

People who want one assistant for personal context and for reminders, with everything kept in documents they own. Skip it if you want a bot that chats with other people, since the chat panel belongs to the n8n editor.

You need an OpenAI key, a Google account with two Docs, and a Telegram bot if you want the copies.

Where to use it

You chat from the n8n editor, and Telegram shows each answer on your phone as a copy. Making the chat public in the trigger’s options lets people use it outside the editor.

• n8n Cloud or your own n8n: import the file and open the chat panel
• Telegram: a copy of each answer goes to the chat id you set
• A different model: change the model in OpenAI Chat Model
• A different split of memory and notes: edit the System Message in AI Tools Agent

Setup

Download saves a zip with workflow.json and Mindket’s setup notes in your downloads folder. Create the two Docs first, because the workflow needs their ids.

1. In n8n, create a workflow, open its menu, choose Import from File and pick workflow.json.
2. Create two empty Google Docs. Add a Google Docs credential and select it in the four Google Docs nodes.
3. In Save Long Term Memories and Retrieve Long Term Memories, replace <YOUR_MEMORIES_DOC_ID> with the id or address of the memories Doc.
4. In Save Notes and Retrieve Notes, replace <YOUR_NOTES_DOC_ID> with the id or address of the notes Doc.
5. Add an OpenAI credential and select it in OpenAI Chat Model.
6. Add a Telegram credential from your BotFather token and select it in Telegram Response. Replace <YOUR_TELEGRAM_CHAT_ID> with your chat id, or delete the node if you do not want the copies.
7. Open the chat panel and send a message.

How to use it

Talk to the agent as you would to an assistant. When you share a preference, a habit or an important date, it adds a short memory. When you ask it to keep a reminder or an instruction, it adds a note. Open either Doc to read or tidy what it saved.

Example

You write that you are learning Spanish and ask the agent to remind you to call Bo on Friday. The agent saves the first as a memory and the second as a note, each with its date.

Next week you ask what is on your list. The workflow reads both Docs first, so the answer mentions the call and suggests a Spanish phrase for it.

For newcomers and experts

If you are new. n8n draws a workflow as boxes joined by lines, and each box, a node, does one job. An AI agent is a model that can use tools, here two tools that write to Google Docs. A memory is a fact about you, and a note is something to do or keep.

For experts. Merge appends the two Docs, and Aggregate puts them under data, so the prompt reads the memories at the first position and the notes at the second. Window Buffer Memory keeps the last fifty messages of the chat session. The agent’s second output has nothing attached.

Good to know

The original left the model unset, and its one Google Doc placeholder stood for two different Docs.

• Models: the connected OpenAI node had no model set, so it ran on the node’s default gpt-5-mini, although the node was named gpt-4o-mini. OpenAI’s deprecations page, read on 12 November 2026, lists gpt-5-mini as shutting down on 11 December 2026 with gpt-5.6-terra as the replacement. Mindket’s copy names gpt-5.6-terra and renames the node OpenAI Chat Model
• A spare DeepSeek node: the original also held an unconnected DeepSeek node on deepseek-chat, which DeepSeek ends on 24 July 2026 according to its updates page, read on 12 November 2026. It is now DeepSeek Flash Chat on deepseek-flash and still not connected. To use it, add an OpenAI credential with your DeepSeek key and the Base URL https://api.deepseek.com
• Two Docs: the memories and the notes need separate Docs, because the prompt keeps them apart. The original used one placeholder for all four nodes, and Mindket’s copy has one for each Doc
• Telegram: the chat id was an example number, and it is now <YOUR_TELEGRAM_CHAT_ID>. Replies go out in HTML mode, and Telegram’s documentation asks for <, > and & to be escaped, so an answer that holds them can fail to send
• Not run against OpenAI, Google or Telegram: Mindket ran the flow with stand-ins and saw both Docs reach the agent and the answer reach the chat and Telegram
• Dates: the workflow’s timezone is America/Vancouver, so saved entries are dated in that time unless you change it in the workflow settings
• Your data: your messages and both Docs go to OpenAI, the copies go to Telegram and the Docs stay in your Google account
• Credentials: the JSON holds none, so add the OpenAI, Google Docs and Telegram credentials yourself