需自行准备 AI 提供商账号或 API key,OpenAI、Groq 等需密钥;使用本地模型需额外配置 LiteLLM 代理与 Ollama;仅提供 macOS 与 Windows 版本,未提及 Linux 支持
browser-operator-core
BrowserOperatorBSD-3-Clause更新放缓内置多智能体平台的开源 AI 浏览器,可替代 ChatGPT Atlas、Perplexity Comet 等产品,让智能体直接在浏览器里完成网页操作与任务编排。
Browser Operator - The AI browser with built in Multi-Agent platform! Open source alternative to ChatGPT Atlas, Perplexity Comet, Dia and Microsoft CoPilot Edge Browser
从 GitHub Releases 页面下载 macOS 或 Windows 安装包;安装后在 Settings 中选择 AI 提供商、填入凭据并选择模型保存macOS 10.15+ 或 Windows 10 64 位+;8GB 内存(推荐 16GB),2GB 可用磁盘空间。README 未提及 GPU 或显存要求。
开源隐私优先的 AI 浏览器,内置多智能体平台,本地处理网页研究与自动化任务
内置多智能体平台的开源 AI 浏览器,可替代 ChatGPT Atlas、Perplexity Comet 等产品,让智能体直接在浏览器里完成网页操作与任务编排。
核心能力
- 多智能体协作自主完成复杂网页任务
- 所有处理在本地完成,可配合 Ollama 离线运行
- 通过 LiteLLM 兼容 100+ 模型(Claude、GPT、Gemini、Llama)
- 支持 OpenRouter、OpenAI、Groq 等多家 AI 提供商接入
容易踩的坑
- v0.6.0最新正式版2025-12-11 · 2025-12-11 06:10 UTC
Browser Operator Core v0.6.0 Release Notes
🎯 Highlights
Browser Operator Core v0.6.0 introduces persistent user memory and universal LLM compatibility, transforming AI agents from stateless assistants into intelligent partners that remember, learn, and adapt across sessions.
✨ New Features
Cognitive Memory System
- Persistent Agent Memory: AI agents now remember key information across sessions, building accumulated knowledge over time
- Automatic Memory Extraction: Important details automatically captured and indexed from conversations
- Searchable Memory Blocks: Natural language search across all stored memories
- Memory Management UI: Full control over what agents remember with privacy-first design
https://github.com/user-attachments/assets/6ba547b9-d54d-4224-9235-38d80b7cd231
Universal LLM Platform
- 7+ Provider Support: Native integration with Cerebras, Anthropic, Google AI, OpenAI, and any OpenAI-compatible endpoints
- Custom Provider Registration: Connect to internal LLMs, Ollama, vLLM, or any compatible API
- Dynamic Model Discovery: Automatically detect and list available models per provider
https://github.com/user-attachments/assets/9d8f77de-8a58-4bb4-aced-4df28da5e5c8
Conversation Intelligence
- Persistent History: All conversations automatically saved with IndexedDB backend
- Quick Resume: Continue any previous conversation with full context
Streamlined User Experience
- Multi-Step Onboarding: Guided setup wizard for initial configuration
- Panel Visibility Controls: Hide/show DevTools UI panels
What's Changed
- Fix the agent runner bug by @tysonthomas9 in https://github.com/BrowserOperator/browser-operator-core/pull/67
- Cleanup eval-server and renamed to agent-server by @olesho in https://github.com/BrowserOperator/browser-operator-core/pull/62
- fix LiteLLM model issue by @tysonthomas9 in https://github.com/BrowserOperator/browser-operator-core/pull/68
- Added OpenAI compatible "input" field support in /v1/responses API by @olesho in https://github.com/BrowserOperator/browser-operator-core/pull/69
- Refactored Settings UI by @olesho in https://github.com/BrowserOperator/browser-operator-core/pull/65
- Feat/eval js by @olesho in https://github.com/BrowserOperator/browser-operator-core/pull/70
- Feat/capture iframes by @olesho in https://github.com/BrowserOperator/browser-operator-core/pull/72
- Add support for more providers and conversation history by @tysonthomas9 in https://github.com/BrowserOperator/browser-operator-core/pull/71
- Memory and Onboarding Updates by @tysonthomas9 in https://github.com/BrowserOperator/browser-operator-core/pull/76
- Add ability to show or hide other panels by @tysonthomas9 in https://github.com/BrowserOperator/browser-operator-core/pull/77
Full Changelog: https://github.com/BrowserOperator/browser-operator-core/compare/v0.5.0...v0.6.0
- v0.5.02025-11-08 · 2025-11-08 21:11 UTC
查看更新说明
Browser Operator v0.5.0 Release Notes
This release introduces visual intelligence overlays, comprehensive session management, and structured workflow orchestration - transforming AI agents from black boxes into auditable, manageable business tools.
Key Capabilities
Visual Intelligence Layer
- Real-Time Progress Tracking: Live visualization of multi-step workflows with completion status
- Guided Workflows: Interactive overlays guide users through AI-assisted processes
Intelligent Task Orchestration
- Built-in TODO Management: Hierarchical task tracking for multi-step workflows
- Progress Monitoring: Track completion of complex, multi-step operations
Enhanced File Management
- File Browser Interface: Navigate and manage all generated documents
- Syntax-Highlighted Viewer: Preview code, reports, and documents in-browser
- Export Pipeline: Streamlined artifact extraction for downstream systems
What's Changed
- Fix Config Propagation in Automated Mode by @olesho in https://github.com/BrowserOperator/browser-operator-core/pull/61
- Add Browser Operator Provider and improvements by @tysonthomas9 in https://github.com/BrowserOperator/browser-operator-core/pull/66
Full Changelog: https://github.com/BrowserOperator/browser-operator-core/compare/v0.4.0...v0.5.0
- v0.4.02025-10-14 · 2025-10-14 19:46 UTC
查看更新说明
Browser Operator Core v0.4.0 Release Notes
Browser Operator Core v0.4.0 represents a transformative leap in AI agent capabilities. By introducing a virtual file system and web rendering engine, agents can now handle complex, multi-step workflows that were previously impossible. Agents can manage their own temporary state across long-running tasks, generate rich interactive content like reports and presentations, and create learning materials such as interactive flashcards - all within the browser environment.
✨ New Features
Extended Agent Intelligence & Persistence
- Stateful Long-Running Tasks: Agents can now use the file system to maintain context and temporary states across extended workflows, breaking through previous memory limitations
- Complex Task Orchestration: Multi-step processes that require intermediate data storage and retrieval are now possible
- Persistent Working Memory: Agents can save, retrieve, and update information throughout lengthy operations without losing context
Virtual File System & Web Asset Management
- In-Browser File Storage: New virtual file system enables creation, storage, and management of code files directly in the browser using IndexedDB
- Session-Scoped Storage: Files are properly isolated per DevTools session, ensuring clean workspace separation
- 8 New Development Tools:
- Create, read, update, and delete files
- List and check file existence
- Render web applications in sandboxed iframes
- Retrieve form data and user interactions from rendered apps
Web Application Rendering
- Live Preview Capability: Instantly render HTML/CSS/JavaScript applications within isolated iframes
- Interactive Reports: Agents can generate full HTML/CSS/JavaScript reports with charts, graphs, and interactive elements
- Dynamic Presentations: Create slide decks and presentation materials that can be viewed and navigated in real-time
- Sandboxed Execution: Secure iframe implementation with proper content isolation
- Data Retrieval: Automated collection of user interactions and form submissions from rendered applications
https://github.com/user-attachments/assets/fdc58ed0-aa80-4c94-a6e8-37457ca32bb1
Development Infrastructure
- Dockerized Evaluation Server: New local Docker setup for simplified testing and evaluation workflows
- Improved Development Environment: Streamlined setup process for contributors
- Better Isolation: Container-based approach ensures consistent behavior across different development machines
🔗 Links
- [Full Changelog](https://github.com/BrowserOperator/browser-operator-core/compare/v0.3.4...v0.4.0)
- [Pull Request #55](https://github.com/BrowserOperator/browser-operator-core/pull/55) - Web Asset Rendering & Virtual File System
- [Pull Request #52](https://github.com/BrowserOperator/browser-operator-core/pull/52) - Docker Evaluation Server
Browser Operator Core continues to evolve as a powerful general AI agent platform. We're excited about the possibilities these new features unlock and look forward to your feedback!
- v0.3.42025-09-22 · 2025-09-22 07:00 UTC
查看更新说明
🎉 Major Update: Specialized Search Agent & MCP Enhancement
✨ New Features
Introducing the Specialized Search Agent
A powerful new agent designed for deep research and discovery tasks that regular search can't handle.What It Does:
- Finds niche professional lists (content writers, VC partners, industry experts)
- Aggregates data from multiple sources automatically
- Builds comprehensive databases from scattered information
- Navigates through LinkedIn, directories, and professional networks
Example Use Cases:
- "Find all Series A venture partners in Silicon Valley focused on AI"
- "Create a list of technical content writers with SaaS experience"
- "Compile contact information for sustainability consultants in Europe"
- "Find all podcast hosts in the crypto/web3 space"Introducing the Specialized Search Agent
A powerful new agent designed for deep research and discovery tasks that regular search can't handle.
MCP Connectors Catalog
- Browse and discover available MCP connectors in one place
- Add, enable, and configure connectors with one click OAuth login.
AI-Powered Tool Selection
- LLM automatically selects the best tools for your task
- Expanded agent capabilities with Search, Web Task, and Action variants
🔧 Improvements
Rock-Solid MCP Connections
- Automatic retry logic for failed connections
- Real-time connection status updates
- Improved tool naming and resolution
- Better error handling and recovery
https://github.com/user-attachments/assets/17e389f4-9fcd-4d9b-9b4c-261014bf8198
What's Changed
- MCP OAuth Support and more updates by @tysonthomas9 in https://github.com/BrowserOperator/browser-operator-core/pull/51
Full Changelog: https://github.com/BrowserOperator/browser-operator-core/compare/v0.3.3...v0.3.4
- v0.3.32025-09-13 · 2025-09-13 22:55 UTC
查看更新说明
What's New
Model Context Protocol (MCP) Support
- Added support for initial MCP implementation
- Improved integration capabilities with external tools
Enhanced Agent Framework
- Updated agent architecture for better performance
- Improved reliability and extensibility of agent
🐛 Bug Fixes
- OpenRouter OAuth: Fixed authentication issues with OpenRouter
- Model Configuration: Resolved model selection and refresh problems
https://github.com/user-attachments/assets/837abaa1-3487-4370-ac65-6175c32254eb
What's Changed
- Update agent framework structure by @tysonthomas9 in https://github.com/BrowserOperator/browser-operator-core/pull/43
- Fix the model config issue, OpenRouter Oauth issue and model refresh … by @tysonthomas9 in https://github.com/BrowserOperator/browser-operator-core/pull/45
- Feat/automation build by @olesho in https://github.com/BrowserOperator/browser-operator-core/pull/48
- Add initial MCP version with SDK imports by @tysonthomas9 in https://github.com/BrowserOperator/browser-operator-core/pull/44
- fix: build issue with latest main merge by @tysonthomas9 in https://github.com/BrowserOperator/browser-operator-core/pull/49
Full Changelog: https://github.com/BrowserOperator/browser-operator-core/compare/v0.3.2...v0.3.3
用户点评
登录用户的真实使用记录,可按版本和硬件筛选还没有人点评。第一个说出真实体验的人最有价值。

Open-source, privacy-focused AI browser that transforms how you work on the web. An intelligent platform for research, analysis, and automation—running locally in your browser.
🚀 Getting Started
⬇️ Download for macOS | ⬇️ Download for Windows
System Requirements: macOS 10.15+ or Windows 10 (64-bit)+ • 8GB RAM (16GB recommended) • 2GB free disk space
AI Provider Setup
| Provider | Best For | Setup |
|---|---|---|
| OpenRouter | Beginners, 400+ models (Claude, GPT, Gemini, Llama) | Sign in through browser |
| OpenAI | GPT models | API key required |
| Groq | Ultra-fast inference | API key required |
| LiteLLM | Local models, privacy, advanced users | Proxy + Ollama setup |
Quick Setup: Settings → Select provider → Enter credentials → Choose model → Save
🎬 What You Can Do
Multi-Agent Automation – Specialized AI agents work together to handle complex web tasks autonomously
Privacy-First – All processing happens locally on your machine. Use local models with Ollama for complete offline operation
Extensible – Compatible with 100+ AI models through OpenAI, Claude, Gemini, Llama, and more via LiteLLM
Use Cases
Research & Analysis – Literature reviews, data collection, competitive intelligence, market research
https://github.com/user-attachments/assets/225319db-c5a0-4834-9f37-5787fb646d16
Shopping & Price Tracking – Product comparisons, review analysis, price monitoring
https://github.com/user-attachments/assets/c478b18e-0342-400d-98ab-222c93eecd7a
Business Automation – Talent sourcing, lead generation, compliance audits
https://github.com/user-attachments/assets/90150f0e-e8c8-4b53-b6a6-c739f143f4a0
👥 Community & Contributing
Get Help
- 📖 Documentation – Comprehensive guides and API references
- 💬 Discord – Join our community for support and discussions
- 🐛 GitHub Issues – Report bugs or request features
Contribute
- Submit pull requests, report bugs, improve docs, or share ideas
- See Contributing Guide and build instructions in our docs
📄 License
Browser Operator is released under the BSD-3-Clause License.
⭐ Star this repository to support open-source AI browser development