DeepSeek Harness course

DeepSeek Harness: FREE 1 Hour Course — From Install to Agentic OS

🎬 Julian Goldie SEO 📅 Aug 17, 2026 ⏱ 58:43
DeepSeek Harness Creator Mode Cordis agentic OS Obsidian

🧠 Model + Harness = Agent

The one idea that makes the whole course click: "An AI model is a brain — it can think, but it can't touch anything. It can't open files, remember last Tuesday, use a browser, or run a program. A model is a very smart voice in the jar. The harness is everything wrapped around that brain — the hands, memory, workspace, and rules." 2:04

Why harnesses matter more than brains: "A top model might be 5-10% better than the one below it. A good harness makes the same model dramatically more useful — it decides what the model can see, what it can do, and how it recovers when something goes wrong. The industry spent 2 years obsessing over brains; now it's moved to bodies." 2:57

The launch facts: August 13th developer preview, MIT license, 115,000+ GitHub stars (fastest-growing AI project ever), 300+ community plugins in days, led by Tianyi Cui (joined DeepSeek March 2026 from Jane Street), 19-person team from a May proof-of-concept to August release. 0:56

⚡ Level 1: Install & First Task

The install is intentionally friendly — and web-first, unlike every other terminal-only harness: 4:00

StepAction
1Install Node.js (nodejs.org)
2Open terminal (Mac: cmd+space "terminal"; Win: PowerShell)
3One install command → runs locally at localhost:3080, opens in your browser
4Connect a brain — DeepSeek API key (platform.deepseek.com), or OpenRouter/OpenCode for free models
5Add a workspace — one clean folder (the "room" the agent works in)
6Write your first real task in plain English
The web-first choice: "Almost every other harness lives inside the black terminal window, which quietly filters out everyone who isn't an engineer. DeepSeek runs as a normal web page — chat window, sidebar, buttons, folder picker. Software you already know how to use." 5:20

The workspace is a safety + focus boundary: "It can read, create, and change files in that folder — nothing outside it. Don't point it at your whole desktop; make one clean folder. It keeps things safe and keeps the agent focused — digging through thousands of irrelevant files costs tokens and time." 7:40

The prompt pattern — Look / Produce / Rules: "What to look at (read the notes in this folder), what to produce (a one-page summary as summary.md), what rules to follow (simple language, short bullets). You don't need magical prompts — just clear briefs, like you'd message a friend." 10:10

His advice: "Give it real jobs from day one — don't vibe-code a snake game. Look at what you spend time on daily and automate that. One automation per week." 12:20

🔍 Trajectory View

"DeepSeek made every step visible." The trajectory view lays out every single step the agent took, graphically clickable and in order: the initial system prompt, your message, the context loaded, its actual thinking (reasoning in full), every tool call, every command run, the exact payloads and results, and how long each step took. 13:51

Why it matters — trust: "Agents break all the time. They're easily misled, easily distracted. The difference between agents you can trust with real work and agents you can't is whether you can see why a mistake happened. With most tools, a failed job gives you nothing — output's wrong, good luck. With trajectory view, you scroll to the exact step where things went sideways: 'it looked in the wrong folder, misread that line, the file was empty.' Fix that one thing, run again." 15:35

This is "the complete opposite of closed labs, where the model's thinking is hidden from you." You can also export any session as a zip — a complete, permanent log in standard format, "so you can check the work without redoing it. Review the trajectory the way a manager reviews a report." 17:00

Version 0.1 disclaimer: "It's 10% of the full version — a developer preview. Things will break, plugins may break next month. If you wire your whole operation to it this week and treat it as finished software, something's going to break. Build it into a system (level three), not a finished product." 17:50

🧩 Level 2: Everything Is a Plugin

DeepSeek's GitHub headline is three words: everything is a plugin — and they mean it literally. "The sidebar is a plugin. The tools are plugins. The system prompt is a plugin. The sandbox is a plugin. Even the agent loop itself — the core think-act-check-repeat cycle — is a plugin. Every organ of this machine can be switched off, swapped out, or replaced without breaking the rest." 20:26

The sealed-car vs modular-car analogy: "Every other agent product is a sealed car — change the radio, add a phone mount, but the engine is the engine. DeepSeek Harness is a car where the engine, wheels, seats, and steering wheel all pop out and click back in while the car keeps driving. Better memory system ships next month? Swap that one part. Better browser tool? Same. You don't migrate to a new product — you upgrade one piece." 21:30

Under the hood is Cordis — and DeepSeek did what labs almost never do: published an actual research paper on the design. The core idea, simply: "Every action can be cleanly reversed. Add a piece, and there's a guaranteed clean way to remove it — no leftover mess, no side effects. That's the reason 'everything is a plugin' actually works. The swapping is safe by design, mathematically — not held together with tape and hope." (Also "very Jane Street: build the boring, rigorous foundations first so the exciting stuff on top doesn't fall over.") 23:18

The four agent presets: 23:43

ModePurpose
StandardHome base — balanced, good tool judgment, everyday summaries/drafts/research
CodeHeavier building powers for complex multi-step technical jobs
MinimalStrips down to almost nothing — light, fast, no tools (quick questions)
CreatorBuild new plugins by prompting

🛠️ Creator Mode

"Creator mode exists for exactly one job: building new plugins by prompting. Describe a tool you wish existed in one plain English sentence, and the harness builds it live inside your session and switches it on." 25:10

The demo: "add a small panel in the bottom corner that shows my three tasks for today and let me tick them off." The harness loads its own plugin-development skills, builds the panel, presents it for approval, and — one button later — the panel appears. "You didn't write a line of code. You just described what you want, and the thing came into existence in front of you." Then: "make it smaller, move it left, different colors, keep it forever." 26:10
The philosophical flip: "Every piece of software you've ever used was somebody else's guess about what you need — designed for an imaginary average user who isn't you. Creator mode flips that: the interface becomes clay. The gap between 'I wish this tool did X' and the tool doing X collapses to one sentence and one approval click." 29:12

Plugins live in-session by default, but can be persisted to become permanent parts of your harness. You can stop, start, edit, and delete them anytime. 28:10

🗂️ The Hidden Config File

"The buried setting almost nobody will find": on your machine there's a .dsh folder → profiles/webcordis.patch.yml. Open it in any text editor. 30:30

What you can do: "write the ID of any plugin — including core interface pieces like the sidebar — add an indented line disabled: true, save, reload, and that piece is gone. You can edit the harness and mold it however you want — or get the agent to do it for you. That's what they mean by editing the harness." 30:35

Plus: community plugin marketplaces (e.g. deepseekcode.com/plugins) are already appearing, and there's a plugin search/install flow in settings. "It creates almost an app store for DeepSeek — everything is a plugin, every piece can be swapped in and out." (With the standard caution: vet anything you install from GitHub.) 32:10

🔗 MCP, ACP & agents.md

"One line in the docs matters more than the whole launch announcement": DeepSeek Harness supports MCP, the Agent Client Protocol, and reads agents.md / claude.md files. 33:08

StandardWhat it does
MCPA standard plug socket for tools — any agent connects to any tool without custom wiring
Agent Client ProtocolA standard way for agent software to talk to the apps around it
agents.md / claude.mdInstruction files — rules, writing voice, projects, where things live, what to never do
The strategic point: "Claude Code reads them, OpenClaw reads them, Hermes reads them, now DeepSeek reads the exact same files. They could have invented their own format and built a walled garden — instead they built the harness to speak the languages the existing agent world already speaks. A challenger that builds for compatibility instead of locking." Start an agents.md today: "a briefing document — the one you'd hand a new hire." 34:25

🏗️ Level 3: The Agentic OS

The reframe that dissolves tool fatigue: "The picture in your head might be a treadmill — every new tool is another thing you must master, so more tools means falling behind. But mastery of individual tools was never the skill. The skill is running one system that new tools plug into. When you have the system, a new tool isn't homework — it's a free upgrade delivered to your door." 36:30

Same philosophy, one level up: "Inside one tool, everything is a plugin. The agent OS is the same idea one level up — every tool is a plugin inside one system. And the biggest open lab in the world just validated the architecture." 37:45

His "Mission Control" dashboard sits every agent side by side — SEO workflow, video agent, loop system, kanban, memory, and DeepSeek Harness — all in one place. But the thing that makes it a system rather than parallel apps is shared memory via Obsidian: 38:20

Obsidian's plainness is the superpower: "It's just a folder of plain-text notes. That plainness is why it works — plain text files in a folder are the one thing every single agent can read and write. No special format, no company's database. One agent learns, every agent knows, because they're all looking at the same folder." 38:50

The vault structure is deliberately simple: Daily / Projects / Research / Outputs, with an agents.md at the top. "Complicated vaults don't work. A handful of folders: daily notes, one note per active project with status, research, and output drafts. Every agent that opens the vault gets briefed automatically. If you can organize a filing cabinet, you can build this." 39:10

The payoff: DeepSeek Harness "knows my brand from day one because it's pulling context from Obsidian — what we work on, who we work with, our FAQs." Multiple agents can code into the same folder simultaneously (six conversations building features at once), and the memory updates as agents work — "a positive feedback loop." 42:20

Model choice: V4 Pro ("frontier, fantastic for building") vs V4 Flash ("not as smart, great for volume, cheap"). Plus the free route — OpenCode's free tier offers DeepSeek V4 Flash, or type "free" on OpenRouter to find free models to plug in. "The harness is free, and now the brain is free too." 41:06

✅ Key Takeaways

  1. The harness, not the model, is the work. A model is a voice in the jar; the harness is the body — what it can see, do, and how it recovers.
  2. Everything is a plugin — literally. Even the agent loop and the system prompt. The Cordis runtime makes every change reversible, so swapping parts is mathematically safe.
  3. Creator mode is the killer feature. Describe a tool in one sentence, approve it, and it appears. "The interface becomes clay."
  4. Trajectory view is the trust mechanism. Every step, visible and clickable — you can diagnose exactly where a job went wrong.
  5. Compatibility is a strategy. MCP + ACP + agents.md/claude.md means DeepSeek speaks the language every agent already speaks.
  6. Level three is the real endgame. The skill isn't mastering tools — it's running one system they plug into, with Obsidian as shared plain-text memory.
  7. It's version 0.1. Powerful, but a developer preview — build it into a system, not as a finished dependency.

🔗 Resources & Links

📍 Timestamp Index

0:00 Intro — free AI employee, three levels
0:56 Launch facts (115k stars, MIT, Tianyi Cui)
2:04 Model + harness = agent
4:00 Install (Node.js, one command, localhost:3080)
10:10 Look / Produce / Rules prompt pattern
13:51 Trajectory view explained
17:50 Version 0.1 disclaimer
20:26 Level 2: everything is a plugin
23:43 The four agent presets
25:10 Creator mode live
30:30 The hidden config file (.dsh)
33:08 MCP, ACP & agents.md
36:30 Level 3: the agentic OS
38:20 Obsidian as shared memory
☰ View all