OpenOSINT

Name: OpenOSINT
Author: OpenOSINT

作者：OpenOSINT版本 2.38.0i★ –i在你的电脑上运行

注册名称io.github.OpenOSINT/openosint

安装

概览

作者提供的说明（英文原文）

AI-powered OSINT agent & MCP server. 21 tools: email, breach, IP, WHOIS, DNS, Shodan, GitHub & more.

PopNx MCP 的三大支柱

这三项永远不会合并成一个分数：意见、测量和人气是不同的东西。

用户评价i暂无评价评价功能将在后续阶段与账户功能一起开放
我们的基础检查i1 项中通过 1 项依据公开信息检查；尚未扫描代码
使用情况i尚未统计★ –

它能帮你做什么

我们还没有为它写通俗的介绍。下面是我们从注册中心读取到的事实。

连接方式: 在你的电脑上运行
传输方式: stdio
分类: 聊天与邮件
你需要填写：ANTHROPIC_API_KEY, HIBP_API_KEY, IPINFO_TOKEN, IP2LOCATION_API_KEY, ABUSEIPDB_API_KEY

安装

选择你使用的 AI 应用，然后复制设置。

安装之前

基础检查：1 项中通过 1 项
在你的电脑上运行，拥有你的用户账户的权限（文件、网络、应用）。
需要你自己填写： ANTHROPIC_API_KEY, HIBP_API_KEY, IPINFO_TOKEN, IP2LOCATION_API_KEY, ABUSEIPDB_API_KEY

标为机密的值是你账户的钥匙。千万不要分享给别人，也不要粘贴到聊天里。

你需要

PyPI已安装 Python 的 uv（用于 uvx 命令）

你用的是哪个应用？

ClaudeClaude CodeCursorVS CodeChatGPT

在上方选择你的应用。我们会记住它，下次按钮会显示“安装到 …”。

打开 Claude Desktop → Settings → Developer → Edit Config，把下面的内容粘贴到 claude_desktop_config.json

{
  "mcpServers": {
    "openosint": {
      "command": "uvx",
      "args": [
        "openosint"
      ],
      "env": {
        "ANTHROPIC_API_KEY": "your-value",
        "HIBP_API_KEY": "your-value",
        "IPINFO_TOKEN": "your-value",
        "IP2LOCATION_API_KEY": "your-value",
        "ABUSEIPDB_API_KEY": "your-value"
      }
    }
  }
}

在终端中运行这条命令

claude mcp add openosint --env ANTHROPIC_API_KEY=your-value --env HIBP_API_KEY=your-value --env IPINFO_TOKEN=your-value --env IP2LOCATION_API_KEY=your-value --env ABUSEIPDB_API_KEY=your-value -- uvx openosint

运行前请先填写空着的值。

在 Cursor 中打开

或者自己复制设置

粘贴到项目中的 .cursor/mcp.json，或 ~/.cursor/mcp.json

{
  "mcpServers": {
    "openosint": {
      "command": "uvx",
      "args": [
        "openosint"
      ],
      "env": {
        "ANTHROPIC_API_KEY": "your-value",
        "HIBP_API_KEY": "your-value",
        "IPINFO_TOKEN": "your-value",
        "IP2LOCATION_API_KEY": "your-value",
        "ABUSEIPDB_API_KEY": "your-value"
      }
    }
  }
}

浏览器会先询问是否打开应用，然后应用会显示设置并请你确认。密钥的值会留空，由你自己填写。

在 VS Code 中打开

或者自己复制设置

粘贴到项目中的 .vscode/mcp.json

{
  "servers": {
    "openosint": {
      "type": "stdio",
      "command": "uvx",
      "args": [
        "openosint"
      ],
      "env": {
        "ANTHROPIC_API_KEY": "your-value",
        "HIBP_API_KEY": "your-value",
        "IPINFO_TOKEN": "your-value",
        "IP2LOCATION_API_KEY": "your-value",
        "ABUSEIPDB_API_KEY": "your-value"
      }
    }
  }
}

浏览器会先询问是否打开应用，然后应用会显示设置并请你确认。密钥的值会留空，由你自己填写。

这个服务器在你的电脑上运行，而 ChatGPT 只能连接互联网上的服务器。请选择其他应用。

你的选择通过 Cookie 保存在这台设备上，无需账户。

需要你填写的值

ANTHROPIC_API_KEY 机密必填 env
Anthropic API key — required for the AI agent REPL
HIBP_API_KEY 机密 env
HaveIBeenPwned API key for breach detection (search_breach tool)
IPINFO_TOKEN 机密 env
ipinfo.io token for higher IP lookup rate limits (search_ip tool)
IP2LOCATION_API_KEY 机密 env
IP2Location API key for enhanced IP intelligence (search_ip2location tool)
ABUSEIPDB_API_KEY 机密 env
AbuseIPDB API key for IP abuse reputation checks (search_abuseipdb tool)

这些设置是根据作者在注册中心的条目生成的。如果 README 的写法不同，请以 README 为准。

它能做什么

它在你的电脑上运行；安装后，工具列表会显示在你的 AI 应用中。

基础检查i

我们检查的是任何人都能看到的信息；我们没有阅读代码。不适用的检查项不计入总数。 · 检查于 2026年11月6日

✓ 通过
作者与仓库一致
注册名称已通过 GitHub 账户 OpenOSINT 验证，该账户拥有此仓库
… 尚未检查
仍在维护
我们还没有从 GitHub 读取这个仓库
… 尚未检查
有许可证
我们还没有从 GitHub 读取这个仓库
– 不适用
端点有响应
它在你的电脑上运行，没有互联网端点
– 不适用
使用 OAuth 登录
在你的电脑上运行，改用本地设置
… 尚未检查
工具描述中没有隐藏指令
它在你的电脑上运行；我们只能读取互联网端点的工具列表

与“聊天与邮件”中的其他服务器比较i

星标数多于此分类中其他 1762 个服务器的 0%
发布到注册中心的时间比此分类中其他 1762 个服务器的 95% 更新
基础检查结果等于或优于此分类中其他 1762 个服务器的 100%

与注册中心里这个分类中所有活跃的服务器比较。只有当发布者拥有该仓库时，星标数才计入。

常见问题

OpenOSINT 安全吗？

它在 1 项适用的基础检查中通过了 1 项。我们没有阅读代码。在添加密钥或令牌之前，请先看看这个服务器要求哪些权限，并使用只具备所需权限的密钥。

它需要哪些密钥或代码？

你需要填写：ANTHROPIC_API_KEY, HIBP_API_KEY, IPINFO_TOKEN, IP2LOCATION_API_KEY, ABUSEIPDB_API_KEY

它在我的电脑上运行还是在云端？

它在你的电脑上运行。你的 AI 应用会在需要时启动它，数据不会经过作者的服务器（除非它自己调用在线服务）。

它还在维护吗？

GitHub 仓库最后更新于–，最新版本（2.38.0）于昨天进入注册中心。

哪些应用可以使用它？

支持 MCP 的应用，例如 Claude Desktop、Claude Code、Cursor 和 VS Code。你的电脑上还需要有 Python (uv)。

有什么可以替代它？

同一分类中有：Vexa, emailmd, MCP WhatsApp Cloud

替代选择

同一分类中其他作者的服务器，按星标数排序。

如何看卡片：★ GitHub 星标i三大支柱：评价 · 检查 · 使用情况i更新 = 最新版本发布到注册中心的时间i

Vexa

作者：Vexa-ai

Meeting bot and transcripts for Google Meet, Teams and Zoom. Live or after, speakers labelled.

在云端运行★ 2,949Apache-2.02个月前更新

安装

emailmd

作者：anypost

Render markdown into email-safe HTML, lint drafts for deliverability problems, and preview emails.

电脑或云端★ 1,385MIT30天前更新

安装

MCP WhatsApp Cloud

作者：codespar

MCP server for WhatsApp Cloud API (Meta) — messages, media, and templates via Graph API

在你的电脑上运行★ 277MIT5个月前更新

安装

作者的 README

打开 README

按原文从作者的仓库复制。PopNx MCP 没有审查或背书其内容。HTML、图片和脚本已被移除。链接会离开本站。

mcp-name: io.github.OpenOSINT/openosint

OpenOSINT The OSINT toolkit for AI agents: Agent Skill, MCP server and CLI. Passive by default, no API keys needed to start, and every finding cites the tool call that produced it.

*A real Claude Code session, nothing mocked: install the plugin and run a passive recon on example.com. Sharper MP4.*

Quickstart

Needs uv. No API keys for the default flow.

Claude Code plugin (skills plus a passive MCP server):

/plugin marketplace add OpenOSINT/OpenOSINT

/plugin install openosint@openosint /openosint:osint example.com

Any agent (Codex, Gemini CLI, Cursor and other Agent Skills clients):

npx skills add OpenOSINT/OpenOSINT

MCP server or pip (other clients):

claude mcp add openosint -- uvx --from openosint openosint-mcp

pip install openosint && openosint exposure example.com

OpenOSINT Cloud

The same engine as a hosted chat: ask about a domain, an IP address or a company. Every answer shows which public sources were checked, when, and a source on every claim. No install, no API keys.

Try it: 10 free messages

What you can do

Real output, captured 2026-11-04 on v2.37.0 and trimmed. None of these commands needs an API key or an AI provider.

$ openosint exposure example.com

Grade: B Score: 88/100 Checks: email OK [passive] · registration OK [passive] · tls OK [touches target] · http_headers OK [touches target] [LOW] DMARC has no aggregate report address (dmarc_no_rua) [LOW] Strict-Transport-Security header missing (hsts_missing) [INFO] Domain registered 31.2 years ago (domain_age)

$ openosint dns example.com [DNS] NS: elliott.ns.cloudflare.com., hera.ns.cloudflare.com. [DNS] SPF: v=spf1 -all [DNS] DMARC: "v=DMARC1;p=reject;sp=reject;adkim=s;aspf=s" [DNS] DKIM not checked: selector probing is enumeration (needs --allow-active).

$ openosint rdap example.com [+] Registrar: RESERVED-Internet Assigned Numbers Authority [+] Registered: 1995-08-14T04:00:00Z [+] Expires: 2027-08-13T04:00:00Z [+] Name servers: elliott.ns.cloudflare.com, hera.ns.cloudflare.com

$ openosint ip 8.8.8.8 [+] Hostname: dns.google [+] Org: AS15169 Google LLC [+] City: Mountain View

$ openosint gdelt example.com No geolocated coverage found for 'example.com'. Coverage: the last 15 min (...) $ openosint gdelt ukraine [+] White House, District Of Columbia, United States (38.8951, -77.0364) — 1 article(s)

The same commands are what the skill runs for an agent. All of them are in the CLI reference.

Why OpenOSINT

Real tool executions. Tools are Python code; only their actual output reaches the model.
Provenance. Every finding is shown with the tool call that produced it. AI summaries are leads to verify, not proof.
Passive by default, with noise labels. Tools that touch the target or are noisy are not even offered to the agent until you opt in.
Works with no keys. exposure, dns, rdap, whois, ip, gdelt and dorks need no account.

	Prompt-only OSINT skills	SpiderFoot	OpenOSINT
What runs the lookups	Nothing: the skill is instructions, and the agent improvises	Its own 200+ modules	Python tools shipped with it
Breadth	Whatever the agent can reach	Much broader: 200+ modules and a correlation engine, per its README	A focused tool set
Built to be driven by an agent	Yes, that is the format	Web UI or CLI, per its README	Agent Skill, Claude Code plugin, MCP server, CLI and web UI
Passive/noisy labels, off by default	No	Not claimed here	Yes, per tool
License	Varies	MIT	MIT

SpiderFoot is the better choice when you want a large standalone scanner. OpenOSINT is for agents that need a small set of real, labelled tools.

Tools

Level is the noise label: passive tools are on by default; anything else needs --allow-active. Per-tool docs: openosint.tech/tools.

Tool	Source	Level	Key needed
`search_dns`	dnspython (built-in)	passive	No
`search_rdap`	RDAP (IANA bootstrap)	passive	No
`search_whois`	python-whois	passive	No
`search_ip`	ipinfo.io	passive	Optional
`search_ip2location`	IP2Location API (sponsored)	passive	Required
`search_gdelt_geo`	GDELT 15-minute GKG feed	passive	No
`generate_dorks`	built-in	passive	No
`search_paste`	psbdmp.ws	passive	No
`search_github`	GitHub REST API	passive	Optional
`search_breach`	HaveIBeenPwned API	passive	Required
`search_shodan`	Shodan API	passive	Required
`search_censys`	Censys API	passive	Required
`search_virustotal`	VirusTotal API	passive	Required
`search_abuseipdb`	AbuseIPDB v2 API	passive	Required
`search_dorks_live`	Bright Data SERP API¹	passive	Required
`search_footprint`	Bright Data SERP API¹	passive	Required
`scrape_url`	Bright Data Web Unlocker¹	touches target	Required
`search_username`	sherlock	noisy	No
`search_email`	holehe	noisy	No
`search_domain`	sublist3r	noisy	No
`search_phone`	phoneinfoga	unverified	No

Sherlock, holehe, sublist3r and phoneinfoga are wrapped as local binaries; install them yourself and enable active tools to use them.

Setting up the three Bright Data tools: Using OpenOSINT with Bright Data.

Passive by default

Agents and MCP clients are only offered tools that query third-party data sources or compute locally. Anything that touches the target or sends many requests from your IP is off until you opt in, and is not listed in an MCP client's tools/list at all.

Level	Meaning
passive	Queries third-party sources or computes locally; nothing is sent to the target's own infrastructure (standard DNS goes through your resolver)
touches target	The target's infrastructure can see the request
noisy	High request volume; expect rate limits, blocks and alerts
unverified	Network behavior not yet confirmed by a traffic capture; treated as non-passive
inherits	A fan-out tool that only runs what the current mode allows

Enable active tools with openosint --allow-active …, OPENOSINT_ALLOW_ACTIVE=1, Web UI Settings (loopback only), or openosint-mcp --allow-active. Passive is not private: every service you query sees the query. Per-tool "what it sends" table and the tool-call cap: Safety.

Go deeper

Web UI + globe

openosint web starts a local browser app with streaming chat, tool cards, a news globe and a graph view. Docs.

*An agent investigation landing live on the globe; click a point to pivot. Sharper MP4 · Live demo (bring your own key, no signup).*

Entity graph

A FollowTheMoney entity graph with statement-level provenance and a human review queue for same_as candidates. Docs.

External exposure report (beta)

openosint exposure example.com grades email authentication, TLS, registration and security headers A to F. Docs.

Installation

pip install openosint, or run without installing: uvx openosint web. Python 3.10+. Install guide.

MCP Server

Works in Claude Desktop, Claude Code, Cursor and any MCP client: uvx --from openosint openosint-mcp. Client configs.

Configuration

Keys go in the web UI Settings, ~/.openosint/config.env or the environment; the app starts with none. All variables · CLI reference · Limitations.

Docker

docker compose up --build, then open http://localhost:8080. Docker notes.

Moved from this README: per-tool details are at openosint.tech/tools; CLI flags (old #cli-reference) at /docs/cli; limitations (#limitations) at [/docs/troubleshooting](https://openosint.tech/docs/troubleshooting/

只显示了开头部分；其余内容请在仓库中阅读