← Open WebUI
Creator Magic
One Chat App for All My Agents: Open WebUI
⏱ 82 min video · 3 min read19 May 2026Worth watching
TL;DR
The creator does a live demo of Open WebUI, a locally-hosted chat interface that unifies multiple AI agents (Hermes, OpenClaw), local models (Qwen 3.6, GPT-OSS 120B, DeepSeek V4), web search, automations, and a calendar into one app running entirely on a Mac Studio M3 Ultra. The appeal is a single, polished UI replacing scattered Telegram agent chats with no cloud dependency and no ongoing API costs.
Key points
1
Open WebUI runs as a locally-hosted web app and as a PWA on iPhone, letting you chat with all your AI agents from one interface without any cloud connection
2
Local models including Qwen 3.6 32B, GPT-OSS 120B, and DeepSeek V4 run on a Mac Studio M3 Ultra with 256GB unified RAM at 15-80 tokens/second, matching frontier model quality
3
Built-in automations (scheduled tasks) let you assign recurring jobs to any agent, making the setup agent-framework-agnostic — if Hermes or OpenClaw breaks, swap it out without rebuilding automations
4
Web search can be enabled inside Open WebUI using engines like DuckDuckGo, Perplexity, or Brave, allowing fully local AI models to search the live web for free
5
The creator used Claude Code to install and configure Open WebUI, connect agents, enable web search, and add new models — recommending this approach over reading docs manually
Actionable insights
→
Paste the Open WebUI quick-start URL into Claude Code and ask it to install and configure everything, including web search and agent connections, rather than following docs manually
→
Enable web search in Admin Panel > Settings > Models, then pick a free engine like DuckDuckGo to give fully local models live internet access at zero API cost
→
Use Open WebUI automations as a central cron scheduler so agent-specific scheduled tasks survive if the underlying framework (Hermes, OpenClaw) is replaced
→
Install Open WebUI as a PWA on iPhone via your local Wi-Fi — voice dictation works offline (using a local Whisper model) and the interface is significantly smoother than Telegram for agent chat
→
Chain web search results into agent memory updates manually or via automation: find a relevant study with a local model, then tell your specialist agent to update its memory with the findings
Notable quotes
“We literally have AGI at our fingertips and it's for free. And they can't take it away.”
“I want to be agent framework agnostic because of all the trouble OpenClaw had a few weeks ago where everything just broke... I can just swap out Hermes for the next thing and it will still work inside Open WebUI.”
“I copied that URL, I pasted it into Claude Code, and I said, I want this running on a throwaway container on my local server behind me. And then my Claude Code was just like, sure.”
Worth watching?
✅
Worth watching the full video?
Worth watching if you want to see live demos of the agent-switching, automation scheduling, local web search, and PWA phone UI in action — the key concepts are captured here but the visual walkthrough adds clarity on the UI and workflow.
Saved you some time? The creator still deserves a like.
Want this for your own channels?
Add the channels you follow. Every new video summarised and in your inbox the moment it drops. From £4/month.
Try it free