Oculr MCP
Enables AI agents to create and control real, isolated browser profiles with engine-level fingerprinting and per-profile proxies, providing antidetect browsing capabilities through a Model Context Protocol interface.
README
Oculr MCP — an antidetect browser for AI agents
Oculr is an antidetect browser that AI agents can drive. It exposes a Model Context Protocol server so an agent like Claude can create real, isolated browser profiles, browse and act inside them, and run whole fleets — each profile a real Chromium browser with engine-level fingerprinting and its own proxy, driven over raw Chrome DevTools Protocol with no injected Playwright or Puppeteer globals.
This repository is the connect guide + Claude skill for pointing an MCP client at Oculr. Full docs: https://oculr.ai/developers/mcp
Why an antidetect browser for AI agents
Generic automation stacks announce themselves to every site they touch. Oculr gives an agent the same thing a person has: a real browser with a consistent, engine-level identity that re-applies on every tab, so the agent sees the same site a person would in a stable environment you control. See oculr.ai/ai-agents.
- Real Chromium, native fingerprinting — fingerprinting is compiled into the engine in native C++, not patched on with JavaScript.
- 40+ MCP tools across browsing, profile lifecycle, fleet control and workflow recording.
- Raw CDP underneath — none of the globals automation frameworks add ever touch the page.
- Per-profile proxies with timezone, language and geolocation aligned to each exit.
- Fleet control — one agent call fans out across many region-pinned profiles at once.
- Local-first — profiles and data stay on your machine.
Install
- Download the Oculr desktop app for macOS, Windows or Linux: https://oculr.ai/download
- Open Oculr and use the one-click Add to Claude connect flow, or add the MCP server manually with the command it shows you.
# The desktop app prints the exact command; it looks like:
claude mcp add oculr
Oculr's MCP server speaks local stdio for zero setup, or HTTP with a mandatory bearer token.
Use it
Once connected, just ask your agent:
"Launch the three storefront profiles and check today's listings on each."
The agent calls Oculr's tools — profile_launch, browser_navigate, browser_snapshot,
fleet_run and the rest — to drive real browser profiles. Page structure is compressed
5 to 10x before it reaches the model, so tokens go to decisions instead of raw DOM.
Programmatic clients
- TypeScript / JavaScript:
oculr-sdk - Python:
oculr - Any Playwright, Puppeteer or Selenium script can attach over the standard CDP endpoint.
Links
- Product: https://oculr.ai
- For AI agents: https://oculr.ai/ai-agents
- MCP docs: https://oculr.ai/developers/mcp
- Glossary: https://oculr.ai/glossary
License
MIT — see LICENSE. Use Oculr in compliance with the terms of the sites you automate.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。