Openbiliclaw
OpenBiliClaw 是纯本地、私有、开源的自进化跨平台内容发现 Agent：从跨平台使用、项目反馈与对话中持续深化心理画像，带着对你的理解主动去 B 站、小红书、抖音、YouTube 等来源找内容 / Fully local, private, open-source, self-improving discovery agent that learns from usage, feedback, and dialogue to find content across Bilibili, Xiaohongshu, Douyin, YouTube, and more.
Last verified:
What is Openbiliclaw?
OpenBiliClaw is a pure local, private, open-source self-evolving cross-platform content discovery Agent that runs entirely on your own hardware. It continuously builds and deepens a psychological profile of you through cross-platform usage, project feedback, and conversations, then actively searches Bilibili, Xiaohongshu, Douyin, YouTube, and 30+ other platforms to find content specifically tailored for you.
The tool is designed for users who want personalized content recommendations without their data leaving their device. It understands your preferences across multiple platforms and proactively discovers videos, articles, and content that match your interests. The Agent operates autonomously with a heartbeat mechanism, executing content discovery tasks every 30 minutes without requiring human input.
Key features include cross-platform content understanding, local data privacy, self-evolving recommendations based on your behavior, support for Chinese video platforms (Bilibili, Xiaohongshu, Douyin) that many AI tools miss, and integration with AI models through OpenAI, Claude, Ollama, DeepSeek, and more. The tool is MIT-licensed and completely free to use, with users only paying for AI model APIs if they choose to use cloud-based models.
OpenBiliClaw is ideal for Chinese internet users, content researchers, students learning from video tutorials, competitive researchers tracking product demos, and anyone who wants an AI assistant that actively finds content for them rather than just answering questions. It transforms information consumption from
Openbiliclaw pricing
Pricing model: Freemium
OpenBiliClaw is completely free and open-source under MIT license. There are no paid plans, subscriptions, or freemium tiers. The tool itself costs nothing to download and use. Users only pay for AI model APIs if they choose cloud-based models (OpenAI, Claude, DeepSeek). Alternatively, users can run free local models through Ollama at no additional cost. bibigpt-skill, which adds Bilibili/Xiaohongshu/Douyin support, is also open-source and free, though it requires a valid BibiGPT account with free trial credits available.
Openbiliclaw pros
- Pure local execution - all data stays on your hardware
- Fully open-source with MIT license
- Completely free - no paid tiers or subscriptions
- Self-evolving psychological profile that deepens understanding over time
- Supports Bilibili, Xiaohongshu, Douyin natively
- Cross-platform support for 30+ content platforms
- Autonomous heartbeat mechanism runs every 30 minutes
- Works with multiple LLM providers (OpenAI, Claude, Ollama, DeepSeek)
- Privacy-first - no data sent to external servers
- Actively discovers content instead of passive recommendations
- Claude Code spawn mode integration for complex tasks
- 4000+ community-contributed skills available
- Supports local audio/video file processing
- Structured JSON output for programmatic use
- Async mode handles long videos (2+ hours)
Openbiliclaw cons
- Requires Node.js 18+ installation
- Setup complexity higher than simple chatbots
- Needs browser logged into bilibili.com for some features
- Requires strict guardrails for safe operation
- Cloud LLM APIs cost money if not using local models
- Not all native summarize features work without bibigpt-skill
- Requires dedicated device recommended for production use
- Dangerous commands must be manually blocked
Frequently asked questions about Openbiliclaw
What is OpenBiliClaw?
OpenBiliClaw is a pure local, private, open-source self-evolving cross-platform content discovery Agent. It runs entirely on your hardware, continuously builds a psychological profile of you through usage and conversations, and actively searches Bilibili, Xiaohongshu, Douyin, YouTube, and 30+ platforms to find content specifically for you. Unlike chatbots that answer questions, OpenBiliClaw executes real tasks autonomously.
Is OpenBiliClaw free to use?
Yes, OpenBiliClaw is completely free and open-source under the MIT license. There are no paid plans, subscriptions, or premium tiers. You only pay for AI model APIs if you choose cloud-based providers like OpenAI or Claude. Alternatively, you can use free local models through Ollama at no cost.
What platforms does OpenBiliClaw support?
OpenBiliClaw supports Bilibili, YouTube, Xiaohongshu, Douyin, Xiaoyuzhou podcasts, Ximalaya, and 30+ other content platforms. It also processes local mp3, mp4, m4a, and mov files. With bibigpt-skill integration, it gains full support for Chinese video platforms that native summarizers typically miss.
How does OpenBiliClaw differ from ChatGPT?
ChatGPT answers questions and gives advice like a consultant. OpenBiliClaw reads files, runs commands, modifies code, and executes workflows like a junior engineer with sudo privileges. It has a heartbeat mechanism that executes tasks autonomously every 30 minutes without human input, making it a truly capable assistant that actually does work rather than just chatting.
What LLM providers does OpenBiliClaw support?
OpenBiliClaw is model-agnostic and connects to external LLM providers including OpenAI, Claude, Ollama, DeepSeek, and Gemini. It doesn't sell models itself - it's a scheduler that makes models work. You can configure your preferred provider by setting LLM_PROVIDER and BASE_URL environment variables.
How do I install OpenBiliClaw?
Install OpenBiliClaw with npm install -g openclaw (requires Node.js 18+). Then start the gateway with openclaw gateway start, and configure your AI model with openclaw configure. Visit http://localhost:8800 to start chatting with your AI assistant. The entire setup takes about 3 simple steps.
Is OpenBiliClaw safe to run?
OpenBiliClaw has powerful capabilities but requires strict guardrails to be safe. Recommended safety measures include: running in a Docker container with read-only filesystem, using a dedicated device, blocking dangerous commands (rm, format, dd), specifying writable directories, and reviewing every execution log. If you wouldn't trust an intern with root access to a folder, don't trust the agent there.
How does the heartbeat mechanism work?
The heartbeat mechanism executes tasks autonomously every 30 minutes without human input. You can configure scheduled tasks like running at 8am daily to fetch RSS from subscribed channels, summarize new videos with bibi summarize --json, aggregate into a Markdown digest, and send to Slack or inbox. This transforms information consumption from passive to active automated research at scale.