Uploaded October 2026 | Updated November 2026, 2 days ago
Install OpenCode on Windows and let a free AI coding agent build a working app for you, with no account, no API key and no credit card. Day 6 of the Free AI Setup series: one free AI tool a day, set up step by step.
Watch the whole series in order: youtube.com/playlist?list=PLfUBzZXRNC-k
Previous, day 5, Open WebUI: youtu.be/aZeR_ShmrWg
Next, day 7, llama.cpp: youtu.be/OrmcJ7cLRn8
What you need: Windows, Mac or Linux, Node.js (for npm) and about 200 MB of disk space. No account, no key and no graphics card for the free models. A graphics card is only needed for the optional fully local route with Ollama: Qwen 3.8 27B used about 17 GB of graphics memory in the run. Without a graphics card, stay with the free cloud models.
In this video:
- What OpenCode is: a free, open source coding agent that reads your files, writes code and runs commands in your terminal
- Install OpenCode with npm install -g @opencode/cli and check the version (opencode --version)
- Remove the older OpenCode 1 package (opencode-ai) first, because both use the opencode command
- The deprecated default model, and how to pick a working free model with /models
- Free models with no sign-in: Big Pickle, Nemotron 3 Ultra Free, MiMo V2.6 Flash Free, Step 5 Preview Free (all four answered in under 20 seconds)
- What /connect is for (paid accounts and your own keys)
- Plan mode with Shift+Tab: a full plan, no files changed
- Build mode: a complete to-do web page (345 lines, 8 KB) in about 23 seconds
- /init writes AGENTS.md, the notes every new session reads (OpenCode reads AGENTS.md, not CLAUDE.md)
- A second task: filters and a dark mode switch that remembers your choice, 13 edits in about two minutes
- Fully local: OpenCode finds Ollama by itself, Qwen 3.8 on an RTX 5090 writes and runs a tip calculator, ollama ps shows 100% GPU
- Tips: private code, opencode run for one-shot jobs, /review before you commit
- Claude Code (from $20 a month) vs OpenCode ($0, with free models)
Chapters:
0:00 Intro: a free coding agent builds an app in 23 seconds
0:32 What OpenCode is: a free coding agent
0:59 What you need
1:27 Step 1: Node.js
1:48 Step 2: Install OpenCode
2:12 Gotcha: remove OpenCode 1 first
2:36 Step 3: Project folder and first start
3:23 Gotcha: the default model is retired
3:44 Step 4: Choose a free model
4:48 What /connect is for
5:12 Step 5: Plan mode
6:11 Step 6: Build mode, a to-do app
6:40 Step 7: /init and AGENTS.md
7:24 Step 8: A second task, filters and dark mode
8:23 Step 9: Fully local with Ollama
9:41 Three tips
10:10 OpenCode vs Claude Code
10:42 Next: llama.cpp
Links
OpenCode: opencode.ai
OpenCode docs: opencode.ai/v2/docs
OpenCode download page: opencode.ai/download
Node.js: nodejs.org
Ollama (day 1 of this series): ollama.com/download
Tomorrow, day 7: llama.cpp on Windows.
New free AI tool every day. Subscribe so you don't miss one: youtube.com/@KGPTalkie?sub_confirmation=1
Install OpenCode on Windows and let a free AI coding agent build a working app for you, with no account, no API key and no credit card. Day 6 of the Free AI Setup series: one free AI tool a day, set up step by step.
Watch the whole series in order: youtube.com/playlist?list=PLfUBzZXRNC-k
Previous, day 5, Open WebUI: youtu.be/aZeR_ShmrWg
Next, day 7, llama.cpp: youtu.be/OrmcJ7cLRn8
What you need: Windows, Mac or Linux, Node.js (for npm) and about 200 MB of disk space. No account, no key and no graphics card for the free models. A graphics card is only needed for the optional fully local route with Ollama: Qwen 3.8 27B used about 17 GB of graphics memory in the run. Without a graphics card, stay with the free cloud models.
In this video:
- What OpenCode is: a free, open source coding agent that reads your files, writes code and runs commands in your terminal
- Install OpenCode with npm install -g @opencode/cli and check the version (opencode --version)
- Remove the older OpenCode 1 package (opencode-ai) first, because both use the opencode command
- The deprecated default model, and how to pick a working free model with /models
- Free models with no sign-in: Big Pickle, Nemotron 3 Ultra Free, MiMo V2.6 Flash Free, Step 5 Preview Free (all four answered in under 20 seconds)
- What /connect is for (paid accounts and your own keys)
- Plan mode with Shift+Tab: a full plan, no files changed
- Build mode: a complete to-do web page (345 lines, 8 KB) in about 23 seconds
- /init writes AGENTS.md, the notes every new session reads (OpenCode reads AGENTS.md, not CLAUDE.md)
- A second task: filters and a dark mode switch that remembers your choice, 13 edits in about two minutes
- Fully local: OpenCode finds Ollama by itself, Qwen 3.8 on an RTX 5090 writes and runs a tip calculator, ollama ps shows 100% GPU
- Tips: private code, opencode run for one-shot jobs, /review before you commit
- Claude Code (from $20 a month) vs OpenCode ($0, with free models)
Chapters:
0:00 Intro: a free coding agent builds an app in 23 seconds
0:32 What OpenCode is: a free coding agent
0:59 What you need
1:27 Step 1: Node.js
1:48 Step 2: Install OpenCode
2:12 Gotcha: remove OpenCode 1 first
2:36 Step 3: Project folder and first start
3:23 Gotcha: the default model is retired
3:44 Step 4: Choose a free model
4:48 What /connect is for
5:12 Step 5: Plan mode
6:11 Step 6: Build mode, a to-do app
6:40 Step 7: /init and AGENTS.md
7:24 Step 8: A second task, filters and dark mode
8:23 Step 9: Fully local with Ollama
9:41 Three tips
10:10 OpenCode vs Claude Code
10:42 Next: llama.cpp
Links
OpenCode: opencode.ai
OpenCode docs: opencode.ai/v2/docs
OpenCode download page: opencode.ai/download
Node.js: nodejs.org
Ollama (day 1 of this series): ollama.com/download
Tomorrow, day 7: llama.cpp on Windows.
New free AI tool every day. Subscribe so you don't miss one: youtube.com/@KGPTalkie?sub_confirmation=1










