OpenBiliClaw is a local-first, open-source AI agent for personalized content discovery. It builds a private profile of your tastes on your own machine, then proactively finds videos, posts and articles you are likely to enjoy across Bilibili, Xiaohongshu, Douyin, YouTube, X, Zhihu, Reddit, Weibo and the open web.
OpenBiliClaw is developed on GitHub, where it has about 3,520 stars, with frequent releases in 2026. Unlike platform recommendation feeds, it works for a single user: everything runs locally, the taste profile stays private, and recommendations come with explanations drawn from your own history across platforms. A browser extension and desktop interface provide recommendation cards, a content library, conversation and a profile view, and an optional Flutter client covers mobile and desktop. It can also plug into the DeepSeek Harness as a plugin, registering agent-bridge tools so an assistant can read recommendations and keep learning while you work. It began as a Bilibili-focused tool, hence the name, but has grown into a general-purpose discovery agent, with experimental Instagram and TikTok sources and 30 days of recommendation history kept in the local content library. Releases are frequent and installation packages are provided for quick setup.
OpenBiliClaw is free and open source under the MIT license. You run it yourself; the only possible cost is the LLM API you choose to power its evaluations.
Power users who are tired of engagement-driven feeds and want a private, personal agent that curates content across many platforms around their actual interests.