Uploaded August 2026 | Updated September 2026, 2 weeks ago
Set up DeepSeek Harness on Windows and run it end to end with Qwen 3.8 27B hosted locally on Ollama, then watch the agent build a full portfolio site and deploy it live to Vercel on its own.
This video covers the complete setup from an empty machine: install Node.js, install DeepSeek Harness with a single npx command, add Ollama as a custom provider, point it at your local Qwen 3.8 27B, and give the agent a real job. You need no DeepSeek API key and no paid subscription, because every token is generated on your own GPU. I also walk through the four agent modes, the three permission levels, and the plugin system that makes DeepSeek Harness different from other coding agents. Everything you need is here, so you can follow along start to finish.
📺 Full playlist: youtube.com/playlist?list=PLUkkjv8z4Ni4
⏱ Chapters:
0:00 Intro
0:58 What You Will Build
1:27 Install Node.js First
1:39 Install DeepSeek Harness with npx
2:33 Open the Dashboard on Port 3080
3:33 Choose Your Workspace Folder
4:01 Standard, PTC, Minimal and Creator Modes
4:49 Read Only, Workspace Write and Full Access
5:18 Slash Commands, Skills and Compact
6:14 Add Ollama as a Custom Provider
6:49 The Ollama Base URL
7:31 API Key and Model ID for Qwen 3.8 27B
8:06 Adding More Local Models
9:28 Why Everything Is a Plugin
10:40 First Chat and GPU Check
11:40 Prompt: Build My Portfolio
12:28 Answering the Agent's Questions
15:03 index.html Gets Created
15:54 Removing Placeholders from the Output
16:39 The Finished Portfolio
18:13 Deploy It to Vercel
19:06 The Vercel Login Flow
20:22 Approving a Command Outside the Workspace
20:54 The Live Public URL
21:18 Token and Context Usage
21:51 Wrap Up
🔗 Resources:
DeepSeek Harness: deepseek.com/harness
Install command: npx @deepseek-ai/dsh web
GitHub repo (MIT licensed): github.com/deepseek-ai/deepseek-harness
Quickstart docs: deepseek-harness.github.io/deepseek-harness/en/guide/quickstart
Node.js: nodejs.org
Ollama: ollama.com
Pull the model: ollama pull qwen3.8:27b
Ollama base URL to paste: http://localhost:11434/v1
Vercel: vercel.com
📺 Watch next:
IBM Granite 4.2 vs Qwen 3.8 27B vs Gemma 4 - Local LLM Benchmark: youtube.com/watch?v=tGkpU2V6QKQ
🎓 Go deeper, my Udemy course:
Master Langchain v1 and Ollama - Chatbot, RAG and AI Agents: kgptalkie.com/langchain
Hit like if this got your setup working, and tell me in the comments which model you connected and where you got stuck, because I read every one. Subscribe and turn on the bell for more local AI builds.
#DeepSeekHarness #Qwen #Ollama #LocalLLM #AICoding
Set up DeepSeek Harness on Windows and run it end to end with Qwen 3.8 27B hosted locally on Ollama, then watch the agent build a full portfolio site and deploy it live to Vercel on its own.
This video covers the complete setup from an empty machine: install Node.js, install DeepSeek Harness with a single npx command, add Ollama as a custom provider, point it at your local Qwen 3.8 27B, and give the agent a real job. You need no DeepSeek API key and no paid subscription, because every token is generated on your own GPU. I also walk through the four agent modes, the three permission levels, and the plugin system that makes DeepSeek Harness different from other coding agents. Everything you need is here, so you can follow along start to finish.
📺 Full playlist: youtube.com/playlist?list=PLUkkjv8z4Ni4
⏱ Chapters:
0:00 Intro
0:58 What You Will Build
1:27 Install Node.js First
1:39 Install DeepSeek Harness with npx
2:33 Open the Dashboard on Port 3080
3:33 Choose Your Workspace Folder
4:01 Standard, PTC, Minimal and Creator Modes
4:49 Read Only, Workspace Write and Full Access
5:18 Slash Commands, Skills and Compact
6:14 Add Ollama as a Custom Provider
6:49 The Ollama Base URL
7:31 API Key and Model ID for Qwen 3.8 27B
8:06 Adding More Local Models
9:28 Why Everything Is a Plugin
10:40 First Chat and GPU Check
11:40 Prompt: Build My Portfolio
12:28 Answering the Agent's Questions
15:03 index.html Gets Created
15:54 Removing Placeholders from the Output
16:39 The Finished Portfolio
18:13 Deploy It to Vercel
19:06 The Vercel Login Flow
20:22 Approving a Command Outside the Workspace
20:54 The Live Public URL
21:18 Token and Context Usage
21:51 Wrap Up
🔗 Resources:
DeepSeek Harness: deepseek.com/harness
Install command: npx @deepseek-ai/dsh web
GitHub repo (MIT licensed): github.com/deepseek-ai/deepseek-harness
Quickstart docs: deepseek-harness.github.io/deepseek-harness/en/guide/quickstart
Node.js: nodejs.org
Ollama: ollama.com
Pull the model: ollama pull qwen3.8:27b
Ollama base URL to paste: http://localhost:11434/v1
Vercel: vercel.com
📺 Watch next:
IBM Granite 4.2 vs Qwen 3.8 27B vs Gemma 4 - Local LLM Benchmark: youtube.com/watch?v=tGkpU2V6QKQ
🎓 Go deeper, my Udemy course:
Master Langchain v1 and Ollama - Chatbot, RAG and AI Agents: kgptalkie.com/langchain
Hit like if this got your setup working, and tell me in the comments which model you connected and where you got stuck, because I read every one. Subscribe and turn on the bell for more local AI builds.
#DeepSeekHarness #Qwen #Ollama #LocalLLM #AICoding
