# Nod AI Agent Is WILD: It Refused to Cancel My Subscription (Honest Review) **URL:** \ **Category:** Content Creators\ **Tags:** humans-in-the-loop, prompt-engineer-48, chatbot, agents, running-locally, tech\ **Created:** [13 November 2026 22:30 UTC](https://www.artofsm.art/t/nod-ai-agent-is-wild-it-refused-to-cancel-my-subscription-honest-review/28403 "2026-11-13T22:30:42Z")\ **Posts on this page:** 2\ **Page:** 1
**Author:** ![merefield](https://www.artofsm.art/user_avatar/www.artofsm.art/merefield/32/74_2.png) [@merefield](https://www.artofsm.art/u/merefield)\ **Post date:** [13 November 2026 22:30 UTC](https://www.artofsm.art/t/nod-ai-agent-is-wild-it-refused-to-cancel-my-subscription-honest-review/28403/1 "2026-11-13T22:30:42Z")
[![](https://www.artofsm.art/uploads/default/original/3X/4/4/44a5c590d08566ea71d41510992ebe77b25340d6.jpeg "Nod AI Agent Is WILD: It Refused to Cancel My Subscription (Honest Review)") ](https://www.youtube.com/watch?v=8FzVflVA_8Y) The video reviews Nod, an open-source AI personal agent running locally on Windows 11, highlighting its capabilities in tasks like email triage and subscription management while noting limitations such as incomplete action execution, performance issues with local models, and the need for cautious use with approval gates and human oversight. Although promising for complex workflows and N8N integration, Nod remains a research preview requiring further testing and careful handling of sensitive data. ---
**Author:** ![merefield](https://www.artofsm.art/user_avatar/www.artofsm.art/merefield/32/74_2.png) [@merefield](https://www.artofsm.art/u/merefield)\ **Post date:** [13 November 2026 22:54 UTC](https://www.artofsm.art/t/nod-ai-agent-is-wild-it-refused-to-cancel-my-subscription-honest-review/28403/3 "2026-11-13T22:54:47Z")
The video provides an in-depth review of Nod, an AI personal agent developed by N8N Labs, which operates locally on Windows 11 using WSL2 and Docker containers. Nod is designed as a chat interface backed by an agent with its own workspace, browser, and database, capable of handling tasks like email triage, invoice follow-ups, and subscription cancellations. The reviewer confirms several claims by the developers, including Nod’s open-source Apache 2.0 license, local operation, and multi-platform preview availability. However, some features, such as the need for a ChatGPT subscription or API keys and the approval gate for consequential actions, were only partially tested or remain unproven. Installation involves downloading the Node Setup executable from GitHub, verifying its integrity, and running it on a Windows 11 system with virtualization enabled. Nod creates a Linux distro and pulls several container images, requiring significant disk space and system resources. The reviewer used an RTX 4060 laptop with 16 GB RAM and noted that while cloud models do not heavily rely on GPU, local models demand substantial memory and processing power. The setup process took about ten minutes, mostly due to download time, and the interface allows users to connect various model providers like Open Router or OpenAI. The reviewer conducted multiple demos to test Nod’s capabilities. In email triage, Nod successfully summarized pasted client emails and identified overdue invoices without guessing missing information, demonstrating honesty and accuracy. However, Nod did not perform any actions like sending emails or scheduling reminders because no mail account was connected, highlighting a limitation in the “doing” part of the agent’s functionality. In a subscription cancellation test, Nod correctly assessed refund eligibility based on policy and usage but stopped short of executing any cancellation steps, reflecting its cautious design. Security and reliability were also examined through an injection test involving a hidden command in a fake email. Nod did not execute the malicious instruction, but the test was inconclusive since Nod never actually read the email content. The reviewer noted some bugs, such as improper file naming in the Linux workspace, and emphasized that Nod’s approval system is a speed bump rather than a foolproof barrier against errors or attacks. Local model performance was poor on the reviewer’s hardware, with a single tool call taking 48 minutes and consuming excessive resources, making local use impractical without more powerful GPUs or smaller context windows. In conclusion, Nod shows promise as an honest and cautious AI agent capable of handling complex workflows and integrating with N8N automation, but it remains a research preview with several unproven or incomplete features. The reviewer advises users to start with fake data, use throwaway accounts, enable all approval prompts, and avoid exposing sensitive information. The combination of a careful model, trust checks, human oversight, and limited account permissions is essential for safe operation. The video ends with an invitation for viewers to request further testing of Nod’s approval gate and N8N integration using a proper test mailbox. ## Useful Links - [Nod AI Agent Official Site](https://nod.n8n.io/) — Direct source for Nod AI agent information, installation, and documentation. - [Nod AI Agent GitHub Repository](https://github.com/n8n-io/nod) — Primary source code, releases, and installation files for Nod AI agent. - [Hermes Agent GitHub Repository](https://github.com/NousResearch/hermes-agent) — Source code and documentation for Hermes Agent, a comparable AI agent discussed in the video. - [OpenClaw GitHub Repository](https://github.com/openclaw/openclaw) — Source code and security documentation for OpenClaw AI agent, relevant for comparison. - [N8N Workflow Automation Platform](https://n8n.io/) — Official site for N8N platform, relevant for understanding Nod’s integration capabilities.