Xiaobei
A social-media customer-acquisition agent for micro and small businesses, controlled via WeChat, that automates content creation, publishing, monitoring, and lead generation.
- Source repo
- TeamWiseFlow/xiaobei
- Stars
- ★ 8.6k
- Last updated
- 3d ago
- License
- NOASSERTION
- Primary language
- Python
- FA score
- 56/100 · Major gaps
At a glance
- How it runs
- Works with
- Platform-specific
- Cost
- Free software; you pay for model usage
- Setup effort
- Medium · a few setup steps
- You'll need
- Typical use
- A solo entrepreneur or small business owner without technical skills who wants to consistently post on WeChat and Xiaohongshu to attract customers, preferring to manage everything via WeChat chat.
- Not a fit if
- Users unwilling to keep a 7x24 always-on machine running.
- Publishers on Xiaohongshu/Douyin without a fixed IP or relay service.
- Source review
- 56/100 · Major gaps
What does this agent do, and when should you use it?
Xiaobei (xiaobei) is a social-media customer-acquisition agent built on OpenClaw, designed for OPC/micro/small businesses. It operates as a team of AI agents: the main agent Xiaobei coordinates tasks, an IT engineer handles maintenance, a content producer creates videos and designs, and a sales customer-service rep manages inquiries. Users interact via WeChat, Feishu, or WeCom, enabling tasks such as writing and publishing WeChat articles, creating Xiaohongshu posts, generating and distributing short videos to platforms like Channels, Douyin, and Xiaohongshu, monitoring and analyzing published content performance, replicating viral videos, gathering intelligence from 18 source types via built-in Smart Search, and identifying potential leads. It includes a VIP Club with tutorials, community access, and an official relay service for fixed-IP requirements. Installation is one-command script for macOS/Linux/Windows, with recommended Alibaba Cloud Bailian Token Plan API keys.
Xiaobei receives instructions via WeChat/Feishu/WeCom and dispatches sub-agents (it-engineer, content-producer, sales-cs) to complete tasks. Concrete operations include: writing and formatting WeChat articles and pushing them; generating Xiaohongshu graphics and posters; producing short videos using video-generation models (happyhorse, doubao-seedance, minimax-H3) and distributing to Channels, Douyin, Xiaohongshu; analyzing viral videos from Douyin/Bilibili/Xiaohongshu and imitating them; monitoring published works and running periodic performance reviews; using Smart Search to query 18 source types (Xiaohongshu, Douyin, Weibo, Zhihu, Bilibili, Twitter, YouTube, etc.) for intelligence; automating browser workflows via forked camoufox-cli skills (login, form-fill, publish, interact); scoring and predicting content effectiveness via content-calibrator; collecting data back via published-track; supporting Xianyu ops and WeCom Moments outreach. Underlying engine is OpenClaw v2026.7.1, using Alibaba Cloud Bailian or Volcano Ark LLM APIs.
- A solo entrepreneur or small business owner without technical skills who wants to consistently post on WeChat and Xiaohongshu to attract customers, preferring to manage everything via WeChat chat.
- A social media operator who needs to quickly replicate viral videos and track performance metrics (reads, likes, shares) across platforms, generating daily review reports.
- A market researcher who wants to set up monitoring on specified sources (e.g., industry news, competitor accounts) and receive automatic summaries.
- A sales team wanting to discover potential customers on social platforms and automatically handle inquiries via the sales-cs AI agent to advance deals.
- A founder or investor-relations officer who needs to generate pitch decks, investor materials, and apply for software copyright or ICP filing with help from the agent.
- A content creator who wants to produce branded promotional videos using AI generation and distribute them across multiple platforms without manual editing.
How do you install or deploy this agent?
Preparation: Sign up for Alibaba Cloud Bailian 'Token Plan' to get AWK_API_KEY (optional MODELSTUDIO_API_KEY for video gen). One-line installation: macOS/Linux run bash -c "$(curl -fsSL https://raw.githubusercontent.com/TeamWiseFlow/xiaobei/master/scripts/install.sh)" (or atomgit variant for China); Windows PowerShell run irm https://raw.githubusercontent.com/TeamWiseFlow/xiaobei/master/scripts/install.ps1 | iex (requires Git Bash installed). The script downloads a ~140MB tarball, then prompts for WeChat QR code scan to link the account. Requirements: recommended Ubuntu 22.04, 2 CPU, 8GB RAM, 40GB disk. After install, data lives in ~/.openclaw/ and program in ~/xiaobei/.
How do you use this agent?
After installation, scan the QR code with WeChat and confirm to start chatting. You can give commands directly in WeChat, e.g., 'Write a WeChat article about X and publish it', 'Analyze this viral video' (provide link), 'Monitor certain sources'. The agent automatically delegates tasks. To re-link a different WeChat account, run ~/xiaobei/bin/openclaw channels login --channel openclaw-weixin. To upgrade, re-run the install script. For debugging, use ~/xiaobei/bin/openclaw gateway run. See docs/ directory for advanced configuration.
What are this agent's strengths and limitations?
- WeChat as the sole interface lowers the entry barrier for non-technical users.
- Multi-agent autonomous collaboration resolves technical issues (e.g., API errors) by spawning the IT engineer sub-agent without user intervention.
- End-to-end loop from content creation, publishing, monitoring, and review covers mainstream Chinese platforms (WeChat, Xiaohongshu, Douyin, Weibo, Zhihu).
- Built-in Smart Search provides free access to 18 source types without requiring API keys.
- VIP Club offers tutorials, community, and an official relay service to overcome fixed-IP limitations, reducing DIY costs.
- Core functionality is tied to WeChat/WeCom binding and Alibaba Cloud or Volcano Ark LLM APIs, creating provider lock-in.
- Some publishing features (e.g., Xiaohongshu, Douyin) require fixed IP addresses not available in typical home/office networks, necessitating VIP relay service or self-hosting.
- Deployment requires a 7x24 always-on machine without personal sensitive data, which may be operationally demanding.
- Installation depends on Git Bash (Windows) and downloads a large anti-detect browser (~557MB), prone to network interruptions.
- Ongoing costs include LLM API fees (e.g., Alibaba Token Plan from ~$5/month) and optional VIP Club membership ($23/year).
- The custom license (NOASSERTION) may restrict commercial use and redistribution.
How does this agent compare with similar options?
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| Xiaobei This agent | 56 · Major gaps | CLIFree + model costs | ★ 8.6k | 3d ago | Python | — |
| Flock Multi-Agent Workbench | 46 · Major gaps | Desktop appFree + model costs | ★ 1.1k | 3mo ago | Rust | OpenAI API · Claude API |
| AIHawk – AI Browser Agent | 59 · Major gaps | MCP serverFree + model costs | ★ 32k | 4d ago | TypeScript | Codex · Claude Code |
| apowerb Agent Platform | 61 · Some gaps | Self-hosted serviceFree + model costs | ★ 148 | 2d ago | Python | OpenAI API · Claude API |
How does FollowAgents rate this agent?
Why each dimension lost points
Evidence shows: install scripts require API keys but do not explicitly follow least privilege; user confirmation exists (e.g., WeChat QR binding) but not detailed; data flow transparency partially covered (storage location) but not full flow; sensitive data handling suggests not placing private files but no encryption; dependency security uses lockfiles and pinned versions but no vulnerability scanning; external effects involve multi-platform publishing but no permission control details; rollback preserves user data but no explicit rollback mechanism; source attribution lists open-source projects but without versions or links. Deductions: lack of detailed least privilege, data flow transparency, and sensitive data handling.
Evidence shows: README and code comments consistent, tests exist but no full test report; dependency availability uses pinned versions and lockfiles but no availability guarantees; failure messages covered in tests but no user-friendly error prompts. Deductions: insufficient detail on dependency availability and failure messages.
Evidence shows: README defines target users (OPC/SMEs) and scenarios (self-media lead generation) but not capability boundaries; trigger precision has clear commands and parameters but no trigger conditions; environment fit supports multiple platforms and systems but no specific limitations. Deductions: insufficient detail on capability boundaries and trigger precision.
Evidence shows: README provides directory structure and install notes but no FAQ; naming stability has version numbers but no naming rules; known limitations partially mentioned (fixed IP) but not comprehensive; license is custom MIT modification but full text not provided; changelog exists but not detailed; maintenance responsibility mentions PR rewards but no clear maintainer. Deductions: missing FAQ and comprehensive known limitations.
Evidence shows: README describes outputs (articles, videos) but no examples; marginal value emphasizes free and open-source but no comparison; cost-benefit provides model fees but no overall cost estimate. Deductions: insufficient output examples and cost-benefit analysis.
Evidence shows: claims (e.g., feature list) lack specific evidence; cross-source verification not provided; fact vs inference separation exists but not explicitly marked. Deductions: claims lack traceability, cross-source verification insufficient.
- Install scripts download and execute remote code; ensure source is trusted.
- Deployment machine advised not to store personal or confidential files, but no encryption provided.
- Multi-platform publishing involves account permissions; authorize carefully.
Recent releases
Latest entries from the project's GitHub Releases. Notes are quoted from upstream as written, not rewritten or tested by FollowAgents.
-
Release notes excerpt
What's Changed * 5.7.4 publish by @bigbrother666sh in https://github.com/TeamWiseFlow/xiaobei/pull/480 Full Changelog: https://github.com/TeamWiseFlow/xiaobei/compare/v5.7.3...v5.7.4View on GitHub ↗ -
Release notes excerpt
What's Changed * V5.7.3 release by @bigbrother666sh in https://github.com/TeamWiseFlow/xiaobei/pull/479 Full Changelog: https://github.com/TeamWiseFlow/xiaobei/compare/v5.7.2...v5.7.3View on GitHub ↗ -
Release notes excerpt
What's Changed * v5.7.2 release by @bigbrother666sh in https://github.com/TeamWiseFlow/xiaobei/pull/477 Full Changelog: https://github.com/TeamWiseFlow/xiaobei/compare/v5.7.1...v5.7.2View on GitHub ↗