# 1 Click Human in The Loop | New Human Review n8n Node ## Метаданные - **Канал:** n8n - **YouTube:** https://www.youtube.com/watch?v=1540pDeGZD0 - **Дата:** 17.02.2026 - **Длительность:** 5:20 - **Просмотры:** 4,721 ## Описание @liammcgarrigle Shows off how to use the recent additions to the review nodes, specifically the human review node for tools. You can now click the plus button (+) on the tool connector to add human in the loop. The review message can be customized with the actual requested parameters as well as AI messages. Upgrade to version 2.8.2 or above to get all features highlighted in the video Links: Human in the loop for tools docs: https://go.n8n.io/xteaas Chat node docs: https://go.n8n.io/EiIyPD How to upgrade on cloud: https://go.n8n.io/FxtyFz How to upgrade on self-hosted: https://go.n8n.io/S4VM3P Sign up for n8n cloud: https://go.n8n.io/ckffhO Get help: https://go.n8n.io/qvM223 Time Stamps: 00:00 - Announcement 00:27 - Adding tool human review in regular chat 01:48 - Using params in approval message 00:47 - How to Enable Feature 01:00 - Connection Examples 02:50 - External approval example 03:30 - Customizing buttons 03:40 - External approval example 04:01 - Chat tool request for more info example 04:43 - Extra coming soon teaser 04:58 - Leave a comment! ## Содержание ### [0:00](https://www.youtube.com/watch?v=1540pDeGZD0) Announcement In a time where AI is being trusted with more and more sensitive operations and data, we at NADN are doubling down on human oversight and control options, you can now add a human review step to tools as easy as clicking the connection and picking your platform of choice. With the human in the loop step, it is impossible for that tool to be used without approval. In this video, I'll walk you through three examples of the new patterns this enables in NAD. Here I ### [0:27](https://www.youtube.com/watch?v=1540pDeGZD0&t=27s) Adding tool human review in regular chat have a simple workflow agent that's able to send emails on my behalf. If I ask it to send an email to someone, it will just go and do it. But I don't know about you, I really don't want an AI to send an email in my name without me signing off on it. So let's add a human review step. It's now as easy as just clicking the plus button on the tools connector and then choosing the platform you're using, which in this case I'm using Nadens's Chatub. So I'll pick chat. then we'll set it so that there's an approve and deny option for most providers. That's all you need. But if you're using ChatHub like I am, then you'll also need to change the response method to chat nodes in the chat trigger for this to work. And with that change, you'll also need a respond to chat node after the AI agent to get the AI's response to the chat. Now, if I ask it to schedule a meeting with Jamie this week, it will ask me to confirm before sending. So, we can see that it works, but it's not really helpful if we can't see what we're supposed to be approving. So, let's customize it. We could add a placeholder for the AI to fill in for us. You do that by using money sign from AI with the descriptive field name. And then the next time it runs, the agent will actually provide a value for that since we used from AI. But I actually advise against doing it like this. It's much better practice to see the actual ### [0:47](https://www.youtube.com/watch?v=1540pDeGZD0&t=47s) How to Enable Feature clicking the plus button on the tools connector and then choosing the platform you're using, which in this case I'm using Nadens's Chatub. So I'll pick chat. then we'll set it so that there's an approve and deny option for most providers. That's all you need. But if ### [1:00](https://www.youtube.com/watch?v=1540pDeGZD0&t=60s) Connection Examples you're using ChatHub like I am, then you'll also need to change the response method to chat nodes in the chat trigger for this to work. And with that change, you'll also need a respond to chat node after the AI agent to get the AI's response to the chat. Now, if I ask it to schedule a meeting with Jamie this week, it will ask me to confirm before sending. So, we can see that it works, but it's not really helpful if we can't see what we're supposed to be approving. So, let's customize it. We could add a placeholder for the AI to fill in for us. You do that by using money sign from AI with the descriptive field name. And then the next time it runs, the agent will actually provide a value for that since we used from AI. But I actually advise against doing it like this. It's much better practice to see the actual values that will be used in the tools rather than having it be provided from the AI separately. That way you know for sure exactly what you're approving. And we made this really easy to do. There's a new money sign tool option that lets us dynamically access the tool details that the agent is sending to the tool. So we can say tool. parameters dot message and it will display the exact value that is being set to the message field in the tool. That means you see the actual parameters that will be used in the tool once you approve it. You can manually write out each of the parameters as you like in this message. But since everything in the curly brackets is JavaScript, I just wrote some code to dynamically display every single parameter that's being sent to the tool. So I don't need to remember to update it if I add or remove parameters in the tool. Now if we try it in ChatHub, we can see it shows us the actual data that will be used. So we are making an informed decision when ### [2:50](https://www.youtube.com/watch?v=1540pDeGZD0&t=170s) External approval example pressing allow. Of course, not all AI agents are chatbots. And the human review also works in situations where the one who needs to review isn't the one that's chatting. Here we have an agent that powers a customer service inbox. It has some tools that doesn't need a human in the loop, like escalating to human support or extending a trial. But we don't want to be giving out refunds without someone actually approving it first. So we'll add in a Slack human review step that sends a review message to a support channel. At the time of recording this, you need to approve the tool in the canvas to actually get the test data to autofill in the editor. Something else I really like is being able to customize the ### [3:30](https://www.youtube.com/watch?v=1540pDeGZD0&t=210s) Customizing buttons buttons. So, instead of just saying approve, we can make this button crystal clear by having it say approve, then the actual amount refund. And I'll fill the ### [3:40](https://www.youtube.com/watch?v=1540pDeGZD0&t=220s) External approval example rest of the details that would be, you know, good to see. Just like anywhere else in NADN, you can also reference fields from any previously run node as well. So now when we test this again in Slack, we'll see the actual options to approve or deny with all of the context that we need. We also brought this human ### [4:01](https://www.youtube.com/watch?v=1540pDeGZD0&t=241s) Chat tool request for more info example in the loop functionality to the regular respond to chat nodes as well. I'm adding a chat tool that is an approval response type. We'll name this get user preference and add a description telling the AI to use this tool to get the user's preference whenever there is two options to choose from. And then if we let AI fill the rest of the fields for us, it can use it to ask us anything that has two options. So if there's any follow-up questions it has that, you know, does [clears throat] just have the two options, it's just a click of a button for us to actually answer it. And if two buttons aren't enough or you need it to be more free form, you can just set this response to free text and leave it more open-ended. The team is ### [4:43](https://www.youtube.com/watch?v=1540pDeGZD0&t=283s) Extra coming soon teaser exploring adding editable params in the human in the loop. So think like instead of just approving or denying an amount, you can actually change it as well before you submit it. Though we are still figuring that out, so there still might be a bit of a wait. This addition ### [4:58](https://www.youtube.com/watch?v=1540pDeGZD0&t=298s) Leave a comment! to N8N to me is really important and it signifies NAND's stance as being a secure and reliable AI solution that you and your business can count on. But we want to hear from the community and users, which is you. Now that it's so easy to add, will you be using more human in the loop? Let me know in the comments. I'll be down there answering questions. --- *Источник: https://ekstraktznaniy.ru/video/15117*