Open-Source DevOps · whiteguo233

OpenBiliClaw

OpenBiliClaw is a local-first AI agent that builds a personalized psychological profile of you, then actively hunts for content across Bilibili, Xiaohongshu, Douyin, YouTube, X, Zhihu, Reddit and the web that matches your deep interests. All data stays on your machine in SQLite; no cloud, no accounts, no tracking.

Talk to DEV.co Discuss Implementation

Source: GitHub — github.com/whiteguo233/OpenBiliClaw

1.2k

GitHub stars

Forks

Python

Primary language

MIT

License (OSI-approved)

Key facts

Objective fields from the source. Values we can't verify are shown as “Unknown” rather than guessed.

Field	Value
Repository	whiteguo233/OpenBiliClaw
Owner	whiteguo233
Primary language	Python
License	MIT — OSI-approved
Stars	1.2k
Forks	71
Open issues	38
Latest release	openbiliclaw-v0.3.160 (2026-07-07)
Last updated	2026-07-07
Source	https://github.com/whiteguo233/OpenBiliClaw

What OpenBiliClaw is

Python-based agent using LLM-driven personality profiling (MBTI, cognitive style, values) and five-layer memory architecture to infer latent interests and proactively search multiple platforms via browser extension (Manifest V3) and local REST API. Embeds content using bge-m3 model; runs on Windows, macOS, Linux; frontend via React web UI at localhost:8420.

Quickstart

Get the OpenBiliClaw source

Clone the repository and explore it locally.

terminalbash

1git clone https://github.com/whiteguo233/OpenBiliClaw.git2cd OpenBiliClaw3# follow the project's README for install & configuration

Need it deployed, integrated, or customized instead? DEV.co ships production installs.

Best use cases

Privacy-conscious content discovery at scale

Users who reject centralized platform recommendation algorithms and want all personal data (viewing history, inferred interests, feedback) retained locally on their machine without any cloud transmission or third-party tracking.

Cross-platform interest synthesis

Users with fragmented interests across multiple platforms (e.g., mechanical keyboards on Bilibili, coffee gear on Xiaohongshu, philosophy on YouTube) who want a single agent that understands them holistically and bridges those silos.

Exploratory serendipity and interest expansion

Users seeking content in adjacent domains they have not yet explored (e.g., architecture aesthetics inferred from watch horology; philosophy inferred from quantum physics interest) rather than passive collaborative filtering refinement.

Implementation considerations

LLM API key (OpenAI default; compatible with other providers) and LLM cost tracking: embeddings, personality inference, and content reasoning each incur API calls; budget scales with session activity.
Platform login sessions: extension requires active login to Bilibili, Xiaohongshu, Douyin, YouTube, X, Zhihu, or Reddit in the same browser; session management and cookie persistence are handled by the extension, not the agent itself.
Vector embedding model (bge-m3 ~1.1 GB): can be downloaded on-demand on first launch or pre-bundled in '-with-embedding' desktop package; requires Ollama for local inference or falls back to API-based embedding if configured.
Database schema and backward compatibility: project is v0.3.x and still in active development (last push 2026-07-07); SQLite schema may evolve; review CHANGELOG before major version jumps.
Browser extension manifest and platform compatibility: Manifest V3 supported on Chrome, Edge, Brave, Arc, Vivaldi, Opera; Firefox support status and timeline not clearly stated in README.

When to avoid it — and what to weigh

Requiring fast, out-of-box discovery without setup — Installation involves downloading a desktop app, backend service, and browser extension, plus platform login and LLM API key configuration. Not suitable for users unwilling to spend 15–30 minutes on setup.
Needing multi-user household personalization — OpenBiliClaw is architected as single-user local-first (one SQLite profile per instance). Multi-user setups require separate backend instances or manual data isolation; not designed for family or organizational shared use.
Strict offline-only requirement — While data stays local, the system requires LLM inference (default: OpenAI API or compatible endpoint). The default vector model (bge-m3) can run locally via Ollama, but initial setup may require internet to download ~1.1 GB model.
Enterprises seeking managed SaaS or compliance audit trails — This is a self-hosted open-source project without formal security certification, audit logging, GDPR DPA, or vendor support. Not suitable for regulated industries or large organizations requiring compliance documentation.

License & commercial use

MIT License (OpenBiliClaw itself). Permits commercial use, modification, and distribution provided attribution is included and the license notice is retained. However, the project may depend on third-party libraries and LLM services (OpenAI, Ollama, etc.) with their own terms.

MIT License permits commercial use of the open-source code without royalties. However: (1) you remain responsible for licensing the LLM (e.g., OpenAI API terms), vector models (bge-m3 licensing), and any bundled dependencies; (2) you must retain the MIT notice; (3) no warranty or support is provided by the author. If commercializing, conduct thorough license compliance review of all dependencies.

DEV.co evaluation signals

Editorial assessment — not user reviews. Directional, with an explicit confidence level.

Signal	Assessment
Maintenance	Active
Documentation	Strong
License clarity	Clear
Deployment complexity	Moderate
DEV.co fit	Good
Assessment confidence	High

Security considerations

Local-first architecture reduces exposure to centralized breaches (SQLite on-disk is user-owned). However: (1) browser extension has access to page DOM and cookies on logged-in platforms; audit its permissions and code before use; (2) LLM API calls transmit content snippets and inferred profile data to external LLM endpoints (unless using local Ollama inference); review LLM privacy terms; (3) SQLite database is unencrypted by default; users handling sensitive data should apply OS-level encryption or file-level protections; (4) no security audit, CVE policy, or formal threat model published; relies on open-source community review.

Alternatives to consider

Feedly + Integromat/Zapier

Cloud-based RSS aggregation with workflow automation; does not build psychological profiles or infer latent interests, but offers managed multi-platform content collection and curated feeds without local setup.

Pocket (by Mozilla) + browser extensions

Saves and recommends articles/videos across platforms; simple, cloud-backed, no profile training required. Less agentic and exploratory than OpenBiliClaw; good for passive curation.

Build your own agent using vector DB + LLM chains; offers full control and privacy but requires substantial engineering effort. Suitable if you want to avoid binaries and integrate into an existing data pipeline.

Software development agency

Build on OpenBiliClaw with DEV.co software developers

Start with the Chrome extension, deploy the backend with our AI-assisted setup guide, then log into your favorite content platform. Full setup in ~20 minutes. All your data stays on your machine.

Talk to DEV.co

Related open-source tools

Surfaced by semantic similarity across the DEV.co open-source index.

openagent openmed opensrc

Related on DEV.co

Explore the category and the services that help you build with it.

Open-Source DevOps AI Agent Development Web Application Development API Development Talk to DEV.co

OpenBiliClaw FAQ

Does OpenBiliClaw phone home or share my data?

No. All user data (viewing history, inferred profile, preferences) stays in a SQLite database on your local machine. The browser extension does not transmit behavior data to a central server. LLM API calls send content snippets and inferred interests to your chosen LLM endpoint (e.g., OpenAI) per your API key; review your LLM provider's privacy terms.

Can I use OpenBiliClaw without paying for an LLM API?

Partially. You can use local LLM inference via Ollama for embeddings (bge-m3). However, personality profiling and recommendation reasoning are optimized for cloud LLMs (OpenAI by default). Using only local models will limit recommendation quality and requires more compute; not officially tested or guaranteed.

What happens if I uninstall or the project is discontinued?

Your SQLite database remains on your machine. You can parse it with standard SQL tools or migrate data to another system. The project is MIT-licensed open-source, so the code is yours to fork and maintain independently if needed.

Is this compatible with my Chromebook or mobile device?

Browser extension works on Chromebook (Manifest V3 support). Desktop backend (currently Windows .exe, macOS .dmg) does not natively run on Chromebook or mobile; however, you can access the mobile web UI (`http://<PC-LAN-IP>:8420/m/`) from any device on the same network and pin it to home screen as a PWA-like experience.

Software developers & web developers for hire

DEV.co is a software development agency delivering custom software development services to companies building on open source. Our software developers and web developers design, integrate, and ship production systems — spanning web development, APIs, AI, data, and cloud. If OpenBiliClaw is part of your open-source devops roadmap, our team can implement, customize, migrate, and maintain it.

Ready for privacy-first discovery?

Start with the Chrome extension, deploy the backend with our AI-assisted setup guide, then log into your favorite content platform. Full setup in ~20 minutes. All your data stays on your machine.

Talk to DEV.co Schedule a Consultation