# Octop

> Self-hosted AI assistant where each user gets a team of specialized agents

- Repository: https://github.com/TencentCloud/Octop
- Category: AI agents
- Stars: 10,743 · Forks: 1,372 · Language: Python · License: MIT
- Experience: Some experience
- Page: https://ir.adyog.com/repos/tencentcloud/octop/

Octop is an AI assistant platform for households and small teams. One process serves a web dashboard, a command line, chat channels (Feishu, DingTalk, QQ, WeChat, Telegram, Discord, WeCom) and scheduled jobs. Conversations, workspaces and credentials stay on your machine, in SQLite by default or PostgreSQL if you prefer.

Each user can have several agents, which Octop calls experts. Every expert has its own workspace, model providers, channels and cron jobs. You can start from a library of expert templates, give an expert one of 16 MBTI personalities, and share setups with other users on the same deployment. An AgentTeams mode, still in beta, lets a coordinator hand multi-step work to several experts.

It is a Python and FastAPI app with a React front end. It combines separate Tencent-built parts for the agent runtime, chat gateway, memory and browser automation. Web, chat and cron all go through one in-process handler, with no external message queue. It also speaks ACP both ways: an IDE can use your agent, and your agent can hand coding work to tools like Claude Code or Codex.

## What it does

- **Multi-user, multi-agent**: One admin account can serve a household or team, with JWT-based user isolation. Each user has several experts with their own workspaces, providers and channels.
- **Many chat channels**: Talk to your agents from the web dashboard, desktop apps, Feishu, DingTalk, QQ, WeChat, Telegram, Discord, WeCom, or an HTTP/SSE/WebSocket API.
- **Memory and knowledge base**: Agent memory is stored with the workspace and moves with it. A knowledge base lets agents answer from your own documents, and corpora can be shared within a deployment.
- **ACP in both directions**: Run `octop acp` so editors like Zed or OpenCode can use your agent. Or let your agent delegate to OpenCode, CodeBuddy, Claude Code or Codex, with permission gates.
- **Browser, terminal and remote desktop**: Headless Chromium sessions for web automation, an AI-assisted shell in the browser, and live remote desktop control from the dashboard.

## What you could build

- Set up a personal assistant that writes weekly reports and organizes notes, keeping its memory in its workspace.
- Give each family member their own agent on one shared home server.
- Connect Feishu, DingTalk or WeCom group chats so a team can route tasks to agents.
- Delegate coding tasks to Claude Code or OpenCode from chat through ACP.
- Schedule daily jobs in natural language, such as pushing a summary at a set time.

## Who it's for

Suits people who want to run their own multi-user assistant at home or in a small team; it helps to be comfortable with a terminal and with setting up model provider keys.

## Get started

Run the one-line installer on macOS or Linux (Windows has PowerShell and cmd installers), then open a new terminal so `octop` is on your PATH. The README mentions `octop init` for the first-run setup and `octop run` to start everything. The README text provided here was cut off, so check it for the rest of the steps.

```
curl -fsSL https://finnie-1258344699.cos.ap-guangzhou.myqcloud.com/octop/install.sh | bash
```

Tags: self-hosted, multi-agent, long-term-memory, acp, chat-channels