Tutorial: Set up your own Octop multi-agent AI assistant station in four steps教程：四步搭起自己的 Octop 多智能体 AI 助手站

This machine requires Python 3.12 and above. Go to the "AI Tools" section of this site to download the Octop 1.0.2b4 source code package (15.7 MB), unzip it into the directory, install the dependencies and start: octop run.

After starting, open the console in the browser and follow the wizard to create the administrator and basic configuration. All data falls in the local ~/.octop/ directory (default is SQLite, can be replaced by PostgreSQL), and the backup is to back up this directory.

Enter the model settings in the console and fill in the model service credentials. Octop supports custom base URLs, so all types of OpenAI-compatible services can be connected; after saving, send a test message on the conversation page to confirm that the access is normal before proceeding.

Remember to look at the settings related to usage and concurrency first: when multiple users + multiple agents are running at the same time, the quota and concurrency upper limit are not configured properly, and it is most likely that "some members keep spinning in circles."

Built-in expert library and expert market: select experts according to scenarios (writing, coding, data analysis, etc.), and you can also enable MBTI personality templates to keep the expression style of each agent consistent, so that when working together in a team, you will not have to change your tone every time you change a conversation.

Start a conversation trial run: let the agent write a weekly report, organize meeting notes, or ask questions about the documents you uploaded - the knowledge base will RAG the local documents, and the answers will include your private information instead of relying on model memory.

In the channel settings, connect any one of Feishu, DingTalk, QQ, WeChat, Telegram, Discord, and Enterprise WeChat to the agent, and you can dispatch tasks in the group with @. The cron in the console supports using natural language to describe "summarizing yesterday's group messages at nine o'clock every morning."

Use ACP for development scenarios: Use octop acp to connect OpenCode/Claude Code (delegating coding tasks with permission confirmation), or directly let AI help you check logs and troubleshoot problems in the interactive terminal in the browser. For system-level operations, it is recommended to open tool approval and shell guardrails first.