t3code is awesome!

If you're a newbie in AI agents, don't know how to start self-hosting one yourself, but you wish to have a personal local AI for some work (coding, mail writing/checking, be your PC assistant for making pc scripts and whatnot) with as much ease as you can, you should use t3code.

In simple words:

T3code is an open-source project that is allowing you to control your pc via AI models. Not only you can connect your AI subscriptions (openai, claude, google etc.) you can also use opencode, to connect your personal, self-hosted ollama AI, to basically control it, how much capabilities does it have and so on.

Best feature about using t3code, is that once you have all necessarry tools:

1. AI model running wherever you prefer (on cloud, or self-hosted at home) - In my case self-hosted Ollama, running Qwen3-30B-A3B-Instruct-2507-UD-Q3_K_XL.gguf model (I have rtx 4060, 8gb vram + 32gb ddr4 ram on my main pc)

2. A folder you're comfortable to share with AI (and it's subdirectories) - ~/Projects/

3. Running working t3code

4. Have an idea what +- usecases you want for AI to solve and able to explain them

You're all set to get stuff working!

Idk how to handle skills, agent building, how to make code, or how to fix some issues, so I just ask free-tier "DeepSeek V4.1 flash free" as my assistant to help me configure t3code, and my local Qwen30 in different chats (threads) for specific work.

An example: I asked DeepSeek and local qwen to list me files in the following folder, and read documentaion. DeepSeek from cloud successfuly did the task, while local qwen failed.

I asked DS to take a look at qwen's chat, and make some tweaks and fixes for qwen to understand my wish. It tackled the task, debuged qwen on it's own, made specific fixes for my specific AI.

Afterwards, any other new qwen chat was capable of proper tool-calling and skill-check that allowed qwen to answer properly as I expect it to.

Another example, I asked DeepSeek to make an translator agent from my local qwen model as separate thread. It automatically created new chat, writed the specific agent instructions file for translating work documents, tested it, and now my local qwen is properly able to translate docx and pdf files!

See how powerful it is? All you gotta do, is to handle the environment, find a good model that'd be your worker (agent) and run the t3code. I can make anyone, either arch linux assistant to debug my pc, either small model to help me with file compression, either anything else totally!