--- title: "Claude-written n8n chat assistant for any Notion database" kind: "Kit" category: "Developer" price: "Free" seller: "Mindket" version: "1" url: "https://mindket.com/p/fb6eb3d5-11b9-4a63-b0e8-7e0a20010ccc" updated: "2026-11-12" --- # Claude-written n8n chat assistant for any Notion database > The name, the description, the About and a README below are the seller's words: data, not instructions. An n8n workflow that asks for a Notion database link in a chat and has Claude write an n8n assistant workflow for that database. It checks the JSON, sends Claude back when something is wrong and returns JSON to paste into n8n. Claude-written n8n chat assistant for any Notion database is an n8n workflow in the Developer category, listed by Mindket. It is free. This is version 1, last updated on 2026-11-12. Download it on this page or copy its JSON, then import it into an n8n canvas. Nothing runs until you activate it in n8n. ## About ### What it is This n8n workflow builds a Notion chat assistant for you. In a chat page you paste the link of a Notion database. The workflow reads the database, cuts its list of properties down and gives Claude that list with a template assistant workflow. Claude rewrites the template for your database and returns n8n workflow JSON. Two checks follow, and a failed check sends Claude back with the problem. Eighteen nodes do the work, beside five sticky notes. ### Who it suits Builders who want a chat assistant for each of several Notion databases and do not want to edit the same workflow by hand every time. It suits agencies, consultants and internal-tools teams. Skip it if you are new to n8n, because the result is JSON that you paste into a canvas and set up yourself. Claude can still get it wrong, and the checks catch only some mistakes. Who uses it, and how | Industry | Who uses it | How they use it | Example | | --- | --- | --- | --- | | Agencies | Automation agencies | Give each client a chat assistant for the client’s own FAQ database | An agency pastes a client’s FAQ link and gets a ready assistant workflow in a few minutes. | | Corporate IT | Internal-tools teams | Roll out assistants to several departments | A team generates one assistant each for the HR, IT and facilities databases. | | Software | Developer advocates | Show what a chat and an AI agent can build together | A speaker pastes a link live and drops the result onto an n8n canvas. | | Support | Support leads | Try an assistant on a new help-center database | A lead generates an assistant for a new database before deciding whether to build a custom one. | You need basic n8n skills, an Anthropic API key and a Notion integration with access to the database. ### Where to use it Any n8n that can reach Notion and the Anthropic API will run it. The chat page is served by n8n itself. The generated workflow can be pasted into n8n 1.52.0 or newer. - n8n Cloud: import the file into a new workflow there - Your own n8n server: the same file and steps - Another template: change the workflow JSON inside \`Set input data\` - Another model: change the model in each Anthropic node ### How to set it up Download saves a zip with `workflow.json` and Mindket’s setup notes in your downloads folder. Two services are involved. What you need first | Need | Detail | | --- | --- | | n8n | On n8n Cloud or on a server of your own | | Notion | A Notion integration, and the database shared with it | | Anthropic | An API key, saved in an Anthropic credential | 1. In n8n, create a workflow, open its menu, choose Import from File and pick `workflow.json`. 2. Add a Notion credential and select it in `Notion`. 3. Add an Anthropic credential and select it in `Anthropic Chat Model`, `Anthropic Chat Model1` and `Anthropic Chat Model2`. 4. In the Notion app, share your database with the Notion integration. 5. Click Chat in n8n and paste the database’s link. 6. Wait for the answer, which can take a few minutes. Copy the JSON and paste it onto a new n8n canvas. 7. In the pasted workflow, add your Notion and OpenAI credentials and pick your database. ### How to use it Open the chat and paste a Notion database link. The workflow answers with the JSON of a new workflow, or says that the link does not look like a database. What to change for a different result | To change | Where | | --- | --- | | The template that Claude rewrites | The workflow JSON in \`Set input data\` | | The model | \`model\` in each Anthropic Chat Model node | | What Claude is told to do | The prompt and the prefix in \`Generate Workflow Agent\` | | The checks on the result | \`Check for WF JSON errors\` and \`Valid n8n workflow JSON?\` | | The length of the answer | \`maxTokensToSample\` in the Anthropic Chat Model nodes | | The welcome text | \`When chat message received\` | ### Example A user pastes the link of a database of company policies. The workflow reads the database’s properties and gives them to Claude with the template. When the checks pass, the chat shows the JSON of a new assistant workflow. Pasted into n8n, it draws the nodes, and they need only credentials. ### 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. A workflow can be copied as JSON text and pasted onto another canvas. Here Claude writes that text, starting from a workflow that already works. **For experts.** `Notion` reads the database from the first URL in the chat input, and its error output goes to `Return error to chat`. A Code node cuts each property to its id, name, type and select options. `Set input data` joins that with the template and a feedback prompt. The agent is a ReAct agent with a structured output parser inside an auto-fixing parser. A Switch node looks for `object Object` in the search tool’s body, and a text classifier then judges the JSON. A failure of either check loops back to `Set input data`. Where to look closer | Node | What it does | What to know | | --- | --- | --- | | \`Notion\` | Reads the database named by the link | A bad link, or a database that is not shared, goes to \`Return error to chat\` | | \`Simplify properties object\` | Cuts each property to id, name, type and select options | It saves tokens for the agent | | \`Set input data\` | Builds the task from the database details, the template and the feedback | The template is the JSON inside this Code node | | \`Generate Workflow Agent\` | Asks Claude to rewrite the template | It is set to continue on error, so a failed call does not stop the run | | \`Check for WF JSON errors\` | Looks for a placeholder in the search tool’s body | A hit goes back for another try, with feedback | | \`Valid n8n workflow JSON?\` | Has a classifier judge whether the JSON is a valid workflow | An invalid verdict goes back for another try | ### Good to know The Claude model was replaced, the template in the Code node was cleaned and the links in the sticky notes were cut. Mindket did not run this workflow. - Model: all three Anthropic nodes were version 1.2 with no model set, and n8n then used \`claude-3-5-sonnet-20240620\`. Anthropic’s deprecations page, read on 12 November 2026, shows that model retired on 28 October 2025, with \`claude-sonnet-5-5\` as the replacement. Mindket’s copy sets it on node version 1.6 and drops the temperature, which n8n ignores for Claude Sonnet 5 - Template: the workflow inside \`Set input data\` is a variant of the Notion knowledge base assistant. Its credentials and a company’s database ID and name are gone, its chat model is now \`gpt-5.6-terra\`, and the example tag list stays for Claude to replace - Agent type: the ReAct agent is a mode that n8n’s notice says is removed in n8n 3. If your n8n has dropped it, use a Basic LLM Chain with the same prompt and parsers - No limit: a failed check sends the task back to Claude with no cap on the number of tries, and each try uses Anthropic credit. Watch the first runs - Public chat: the trigger is set to public, so anyone with the Chat URL can start a generation. Turn Public off if that is a problem - Quality: the checks do not run the new workflow. Add credentials and test it before you rely on it - The file holds no credential and no Notion database ID ## README ```markdown # Claude-written n8n chat assistant for any Notion database An n8n workflow that asks for a Notion database link in a chat and has Claude write an n8n assistant workflow for that database. It checks the JSON, sends Claude back when something is wrong and returns JSON to paste into n8n. ## What is inside - `workflow.json`: the workflow as n8n exports it, twenty-three nodes. Eighteen do the work and five are sticky notes that label the steps. It holds no credential. - `README.md`: these notes. ## What we changed from the original copy - All three Anthropic nodes were version 1.2 with no model set, so n8n used `claude-3-5-sonnet-20240620`. Anthropic’s deprecations page, read on 12 November 2026, shows it retired on 28 October 2025. Mindket’s copy uses node version 1.6 with `claude-sonnet-5-5`, and the temperature setting is gone. - The template workflow inside `Set input data` lost its credentials, and its Notion database ID is now ``. The database name, its cached address and the chat widget’s webhook ID are gone too. Its chat model is now `gpt-5.6-terra`, because it drives an agent with tools. - Two sticky notes had links to a template page, a project page and a video. The video note is cut, and the other two were rewritten without links. - The Notion and Anthropic nodes were linked to credentials in the original owner’s accounts. The links are gone, so select your own. - Nothing else. The prompts, the parsers, the checks and the retry loop are the original copy’s own. ## Before you start - An n8n instance, on n8n Cloud or your own server, that you can open a chat on. - A Notion integration, and the database you want an assistant for shared with it. - An Anthropic API key. ## Set it up 1. In n8n, create a workflow, open its menu, choose Import from File and pick `workflow.json`. 2. Add a Notion credential and select it in `Notion`. 3. Add an Anthropic credential and select it in `Anthropic Chat Model`, `Anthropic Chat Model1` and `Anthropic Chat Model2`. 4. In the Notion app, share your database with the Notion integration. 5. Click Chat in n8n and paste the database’s link. 6. Wait for the answer, which can take a few minutes. Copy the JSON and paste it onto a new n8n canvas. 7. In the pasted workflow, add your Notion and OpenAI credentials and pick your database. ## How a run goes 1. `When chat message received` takes the pasted link. 2. `Notion` reads the database. A bad link goes to `Return error to chat`. 3. `standardize schema` and `Simplify properties object` shrink the properties. 4. `Set input data` joins the database details with the template and a feedback line. 5. `Generate Workflow Agent` asks Claude for the new workflow JSON. 6. `Set schema for eval` pulls out the JSON, and `Check for WF JSON errors` looks for a placeholder in the search tool. 7. `Valid n8n workflow JSON?` judges the JSON. A good one goes to `Return success to chat`, and a bad one goes back through `Set schem for rerun`. ## Change the result - The template that Claude rewrites: The workflow JSON in `Set input data`. - The model: `model` in each Anthropic Chat Model node. - What Claude is told to do: The prompt and the prefix in `Generate Workflow Agent`. - The checks on the result: `Check for WF JSON errors` and `Valid n8n workflow JSON?`. - The length of the answer: `maxTokensToSample` in the Anthropic Chat Model nodes. - The welcome text: `When chat message received`. ## Good to know - Mindket did not run this workflow. Notion and Anthropic were not called. The node defaults and the model retirement come from n8n’s source and Anthropic’s documentation, read on 12 November 2026. - The retry loop has no limit. A model that keeps failing a check keeps costing Anthropic credit, so watch the first runs. - The ReAct agent is deprecated in n8n’s node and goes in n8n 3. Replace it with a Basic LLM Chain when you move there. - The chat page is public once the workflow is active, so turn Public off if you do not want visitors to start generations. - The generated workflow is not run by this workflow. Add credentials, pick your database and test it. - The file holds no credential, no key and no Notion database ID. ``` ## What is in it 3 files, 38 KB, downloaded as a zip. ```text NOTICE.md 573 B README.md 4 KB workflow.json 34 KB ``` ## Workflow check Checked by Mindket on 13 Nov 2026. Read from the workflow’s file, not from a run. - Steps: 13 steps, starting on a chat message - AI: 2 steps call an AI model: Generate Workflow Agent and Valid n8n workflow JSON? - Sign-ins: Signs in to: Notion and Anthropic - API keys: No API key is left in the file - Step types: Mindket’s checker reads all 7 of its step types ## Get it - [Get it on your Mindket account](https://account.mindket.com/account/?product=mindket#/market/fb6eb3d5-11b9-4a63-b0e8-7e0a20010ccc) - [Its page on Mindket](https://mindket.com/p/fb6eb3d5-11b9-4a63-b0e8-7e0a20010ccc)