Tecent Octop (Opensource): Wait, Tencent just ENDED OpenAI Dots & Muse!?
In this video, I'll walk you through Octop, a free, open-source AI assistant from the TencentCloud GitHub organization that you can run on your own machine. I install it from the terminal, start it with the right timezone, go through the setup wizard, and connect a model. Then I build a small release-preparation workflow: a Release Helper expert, a TaskBoard knowledge base with local ONNX embeddings, and a cited answer that I check against the source file. After that I add a Reviewer, create an expert team and test it with a deliberately inaccurate draft, delegate a file-writing task to OpenCode through ACP, schedule a weekday reminder, and finish with a quick Browser AI+ test and my verdict. -- Key Takeaways: 🐙 Octop is a free, MIT-licensed, self-hosted AI assistant that combines specialist experts, your documents, scheduled tasks and coding agents. 💸 The app is free, but you supply the models, so cloud API usage can cost money; Ollama is supported for local models. ⚠️ Self-hosting doesn't make every request local: anything you send to a cloud provider still leaves your machine. 💻 The terminal installer sets up Python for you, and you can set the timezone when starting the app so scheduled tasks run on your clock. 🧠 Experts each have their own workspace, model and prompt files, so one can write while another reviews. 📚 Knowledge bases with local ONNX embeddings let an expert cite your documents, and you can check every answer against the source. 👥 AgentTeams (beta) uses a coordinator plus at least two running members; the inaccurate-draft test shows whether the reviewer catches wrong claims. ⚡ Through ACP, an expert can hand file-writing work to OpenCode, but the coding tool needs its own working setup and models. ⏰ Scheduled tasks support fixed Text reminders or Agent jobs, and Execute Now lets you test them; the Octop process has to stay running. ✅ Verdict: start with one expert, one model and one small document, then add the reviewer or the coding runner when you have a reason to. -- Timestamps: 00:00 Intro 00:49 Pricing & privacy (cloud vs local models) 01:14 Installing Octop 01:52 Starting Octop (set your timezone) 02:12 Setup wizard 02:33 Connecting a model (DeepSeek / Ollama) 03:13 Creating your first expert 04:06 The sample release notes 04:37 Knowledge bases + local ONNX embeddings 05:23 Asking with citations (and checking them) 06:25 Reviewer expert + Release Team 06:59 Team conversation & the inaccurate-draft test 08:08 Delegating to OpenCode via ACP 09:24 Scheduled tasks (Text vs Agent) 10:24 Channels, connectors & Browser AI+ 10:53 Verdict
