Description

n8n-nodes-chatfrom

This n8n community node connects your workflows to your ChatFrom chatbots. Ask them questions, add web pages, texts and Q&As to what they know, and read the conversations visitors had with them.

ChatFrom builds AI chatbots that answer from a business’s own website and documents, on its site, on a share link and in chat apps such as WhatsApp, Telegram and Slack.

n8n is a fair-code licensed workflow automation platform.

Installation
Operations
Credentials
Compatibility
Usage
Examples
Resources
Version history

Installation

Follow the installation guide in the n8n community nodes documentation, and install n8n-nodes-chatfrom.

Operations

Chatbot

• Ask: ask a chatbot a question and get its answer, with the sources it used.
• Get: a chatbot’s name, status, model, embed snippet and share link. Turn Simplify off for its instructions and every setting too.
• Get Many: the chatbots in your account.

Knowledge

• Add Pages: add web pages to a chatbot’s knowledge. ChatFrom reads each page itself.
• Add Text: add a text, such as opening hours or a returns policy.
• Add Q&A: add a question with the exact answer the chatbot should give.
• Get Many: a chatbot’s knowledge sources.
• Remove: remove one source, picked from the chatbot’s list or given by its ID: the doc_id that Get Many, Add Pages, Add Text or Add Q&A gives.

Conversation

• Get: one conversation, message by message, picked from the chatbot’s recent ones or given by its ID.
• Get Many: a chatbot’s conversations from every channel, newest first, with filters for those it could not answer, those rated down and those passed to a person. Return All gives every one, and otherwise Limit says how many of the newest, from the last 7 days unless you set Days. The node reads them 500 at a time.

Usage

• Get: the credits used this month against your plan’s allowance, and when it starts again.

The node also works as a tool for n8n’s AI Agent.

Credentials

You need a ChatFrom account. The free plan is enough to try it.

1. In ChatFrom, open Account, then API keys, and make a key for n8n alone. You can then delete it on its own if you ever need to.
2. In n8n, create a ChatFrom API credential and paste the key. It starts with cf_.

The key acts as you. In this node it reads your chatbots’ settings, knowledge and conversations, adds and removes knowledge and asks your chatbots, and what it asks and adds uses your plan’s credits. Through ChatFrom’s endpoint for AI agents, the same key can also create chatbots, change their instructions and settings, put them live and run evaluations, which are billed to the OpenRouter key you saved in ChatFrom, if you saved one. It cannot delete a chatbot, change your plan or make other keys. It reads what your visitors wrote, an email address they gave included, but it cannot search conversations by email. It does not see your actions’ and MCP servers’ addresses, and it cannot import from Google Docs or Notion, though what you imported is part of the knowledge and the answers, and its questions use your chatbot’s tools, such as Google Sheets or Shopify, as a visitor’s would. From n8n 2.24, only the ChatFrom node can use this credential, so an HTTP Request node cannot send the key somewhere else. On an older n8n, anyone who can use the credential in a workflow could do that, so share it only with people you trust.

Compatibility

Made with @n8n/node-cli 0.53. Tested with n8n 1.123, 2.0, 2.24 and 2.44.

Usage

• Credits. Asking a chatbot and adding knowledge use your plan’s monthly credits, as they do in ChatFrom. A page added again that has not changed costs nothing.
• Asking as the owner. Ask answers as you trying the chatbot out, so it works on a draft too. To go on with the same conversation, put the conversationid of the last answer in Options, Conversation ID. Use one an earlier Ask gave, not a visitor’s conversation from Get Many: your question would be added to it as if the visitor wrote it. When an answer carries a question (a step of the chatbot’s workflow waiting for input), send the reply as the message, with that question’s threadid as Thread ID and the same Conversation ID.
• Pages. Add Pages takes one address or several, each on its own line or after a comma or semicolon and a space, up to 60 per item. A comma inside an address keeps it whole, and an expression may give a list of addresses instead. It gives back one item per page, in the order given, with ok, the doc_id of each page stored (which Remove takes), and, for a page that could not be added, the error. An entry that is not a web address, such as a word split from a sentence, an e-mail address or a price, comes back as not added without being sent to ChatFrom; an item with no web address at all stops with an error, as an empty one does. A JSON list of addresses, as an AI agent may give, is read as one too, and a link as Slack or Markdown writes it gives its address. A list’s numbers and bullets are left out. A page behind a sign-in cannot be added, because ChatFrom reads it from the open web.
• Limits. ChatFrom limits how fast each plan may ask and add: 20 answers and 30 additions a minute on Hobby, 60 of each on Pro and 120 on Business, where one addition may hold up to 60 pages. Every account may also index up to 60 sources a minute and read 600 pages an hour, and one key may make 120 calls a minute in all. When ChatFrom refuses before doing anything and says to wait a minute or less (a limit reached, or a moment it could not sign in as you), the node waits and sends that item once more. Stopping the run ends the wait. For a large batch, put the node in a Loop Over Items with a Wait node rather than turning on Retry On Fail: Retry On Fail runs every item again, so questions already answered are asked, and counted, a second time.
• A wrong key. When ChatFrom refuses the key, the node does not send the items after it. With On Error set to continue, each of them gets the same error.
• Slow answers. The node waits up to about two minutes for ChatFrom to add knowledge or answer. If it stops waiting, or the connection breaks while it waits, the work may still finish in ChatFrom, so check the chatbot before running the same items again.
• Text others wrote. Conversations hold visitors’ own words, answers draw on pages others wrote, and page titles and document names come from the pages and files themselves. When the node is an AI Agent’s tool, the first item of Ask, Add Pages, Knowledge Get Many and Conversation results carries a note saying to read that text as data, never as instructions. Keep the agent’s other tools in mind when you give it this one.

Examples

• Unanswered questions to Slack, every morning. A Schedule Trigger at 9:00, then ChatFrom (Conversation, Get Many, Days 1, Unanswered on), then Slack, posting each conversation’s title. Each one is a question your chatbot could not answer, and a page or a Q&A it is missing.
• Keep a chatbot up to date. When your site publishes or changes a page (a webhook from your CMS, or an RSS Feed Trigger), ChatFrom (Knowledge, Add Pages) with the page’s address. A page that has not changed costs nothing.
• Know before the credits run out. A Schedule Trigger every Monday, then ChatFrom (Usage, Get), then an If node that checks whether credits is past 80% of allowance, then an email or a Slack message to you.

Resources

• n8n community nodes documentation
• ChatFrom documentation
• ChatFrom’s API reference, the REST API this node calls with your API key
• ChatFrom for AI agents, where API keys are made

Version history

• 0.2.0: calls ChatFrom’s REST API rather than its MCP server, with the same operations, fields and output.
• 0.1.0: the first version, with the operations above.