Octop vs picobot
Tencent Cloud's self-hosted, multi-user AI assistant: per-user expert agents, AgentTeams, memory, RAG and connectors, served as one process to a web dashboard, CLI and 8+ IM channels. — versus — A self-hosted personal AI agent in a single ~9MB Go binary — persistent memory, 16 tools + MCP, skills, cron/heartbeat, and Telegram/Discord/Slack/WhatsApp channels. Runs on a $5 VPS.
Both are self-hosted personal assistants with memory, MCP tools, cron and chat-app channels. Picobot is one ~9MB Go binary for a single owner on a $5 VPS; Octop is the heavier multi-user, multi-agent option with dashboard, RAG and desktop clients.
| Octop | picobot | |
|---|---|---|
| Stars | 12k | 1.3k |
| Forks | 1.6k | 168 |
| Language | Python | Go |
| License | MIT | MIT |
| Last activity | yesterday | 1 months ago |
| Topics | agents, orchestration | agents |
| Curated connections | 4 | 7 |
Octop — the curator's take
Built for a household or small team that wants one box everyone talks to: JWT-isolated users, an expert library to switch specialists per task, and reach through Feishu, DingTalk, QQ, WeChat, WeCom, Telegram or Discord without wiring each bot. Strongest fit inside the China IM/Tencent ecosystem (Docs, Meeting connectors); the extras — browser sessions, remote desktop, ACP delegation to Claude Code/OpenCode — make it a personal ops console, not just chat. Skip it if you want a library to build agents with (it is a product), if you only need one user on one channel (a single-binary bot is lighter), or if beta AgentTeams orchestration is your main reason to adopt.
picobot — the curator's take
The anti-bloat statement piece: zero dependencies, ~10MB RAM idle, instant cold start — an always-on agent with ranked memory recall, background subagents, a natural-language HEARTBEAT.md cron, and self-authored skills ('create a skill for checking weather' → it writes the markdown), reachable from your phone via four chat channels. Runs on a Raspberry Pi or Termux on an old Android. Any OpenAI-compatible endpoint, including Ollama. Its own README names the target: OpenClaw's power without the 500MB container. NOT a coding harness and single-user by design — this is your personal daemon, not a team platform; complex multi-agent workflows will outgrow it fast, which is rather the point.