The video explains how to significantly reduce the costs of running AI agents like OpenClaw by using HPC AI’s direct API access to high-performance open-source models, bypassing expensive third-party routing fees. It provides a step-by-step tutorial on setting up a VPS, installing OpenClaw, and integrating HPC AI’s API, resulting in substantial savings and efficient AI agent performance.
The video discusses how OpenCore, a popular open-source AI agent, can be expensive to use due to API costs associated with premium models like Claude and GPT. Many users spend significant amounts monthly on these API credits. Even alternatives like Open Router, which use open-source models, still incur extra fees because they route requests through third-party GPUs, adding routing and platform fees on top of compute costs. The video aims to show viewers how to drastically reduce these costs by cutting out the middleman entirely.
The solution presented involves using HPC AI, a company that owns and operates its own GPU clusters and offers direct access to frontier open-source AI models through their API service. Models like Miniax M2.5 and Kimmy K2.5 are highlighted for their high performance and multimodal capabilities. By accessing these models directly from HPC AI, users can save between 53% to 70% on costs compared to other providers, making it a highly cost-effective option for running AI agents.
The tutorial portion guides viewers through setting up an account with HPC AI, obtaining an API key, and preparing a VPS (Virtual Private Server) to run OpenClaw, an AI agent framework. The presenter recommends using a KVM VPS with Ubuntu for ease of setup and affordability, noting that while the tutorial focuses on free usage, a VPS can be obtained for as low as $8 per month if a personal Mac Mini or MacBook is not available. The VPS setup includes upgrading the server, installing Node.js, and then installing OpenClaw itself.
Once OpenClaw is installed, the video walks through configuring the agent, including bypassing certain setup steps, selecting free services like DuckDuckGo for search, and finally integrating the HPC AI API key to connect the agent to the HPC models. The presenter demonstrates how to approve device pairing requests and restart the OpenClaw service to ensure everything is running smoothly. They also show how to switch between different AI models within OpenClaw and test the agent’s responsiveness.
In conclusion, the video emphasizes the significant cost savings achieved by using HPC AI’s direct API access combined with OpenClaw, with the presenter sharing personal usage data showing only a few cents spent after several hours of interaction. The approach offers a scalable, affordable way to run powerful AI agents without incurring high monthly fees. The video ends by encouraging viewers to try the setup themselves and hints at more models becoming available in the future, promising even greater cost-effectiveness.