DeepSeek Harness Tutorial: Free AI Agent Setup [Desktop App]

Every prompt is in my free AI newsletter, ready to copy and paste, with the hotel file I used: https://rakhulkarthick.substack.com/p/deepseek-harness-tutorial-free-ai A DeepSeek Harness tutorial on the free desktop app, with no terminal. DeepSeek Harness is a free, open source AI agent, and you can even run it free on your own laptop. First how it works, in 9 short parts, then 4 real jobs you can copy: a report from a real file, a small app, a team of agents and a plugin it builds just because you ask. What's in the video: - How it works: what a harness is, the install, the screen, the model and what it costs (and running it free with Ollama), your data, permissions, the 4 modes, the Trajectory view, and AGENTS.md and skills. - 4 jobs: a money answer and a Word report from 119,390 real hotel bookings, an invoice app, a 3-agent team that checks every number, and a spend meter plugin. - The catches: its plugin passed its own tests and was still wrong, a stopped job woke up and saved over my report, and the Claude Code and Codex handoffs failed on my Mac. Everything I ran used about 62 US cents of the free credit you get when you sign up. Chapters: 00:00 How DeepSeek Harness works, and 4 real jobs 01:14 1. What a harness is 02:06 2. Install the desktop app 03:39 3. Find your way around 04:43 4. The model and what it costs (free with Ollama) 08:17 5. Where your data goes 09:00 6. Permissions 09:44 7. The 4 modes 11:02 8. Follow one question through the machine 12:48 9. How it knows you 13:38 Job 1: Ask a real file a real question 16:27 Job 2: Build a small app 17:54 Job 3: Give it a team 19:16 Job 4: Build your own plugin 22:18 Is it really a free Claude Code? 23:34 My verdict and the final receipt Links from the video: DeepSeek Harness: https://www.deepseek.com/en/harness GitHub: https://github.com/deepseek-ai/deepseek-harness DeepSeek prices: https://api-docs.deepseek.com/quick_start/pricing Ollama: https://ollama.com Data: Antonio, N., Almeida, A. and Nunes, L. (2019). Hotel booking demand datasets. Data in Brief 22, 41-49. CC BY 4.0. Cleaned copy from TidyTuesday: https://github.com/rfordatascience/tidytuesday/tree/main/data/2020/2020-02-11 I'm Rakhul. Every week I show you how to use AI on real work, with the exact prompts you can copy and paste. If that's useful, subscribe. Rakhul