Instagram Public MCP
Provides tools to fetch public Instagram profile metadata and recent posts without authentication, using an anonymous Instaloader patch with fallback to mobile feed endpoints.
README
Instagram Public MCP for Vercel
A remote MCP server written in Python that queries public Instagram data only without an Instagram login, cookies, passwords, or saved account sessions.
Important: mobile-feed Instaloader patch
This build does not use the stock Instaloader 4.15.3 package.
It pins the anonymous profile/post fix from Instaloader pull request #2722 at:
3f72b47466cc505b92c6e9b0ca2f5689e93a6a89
Dependency:
instaloader @ https://github.com/alimony/instaloader/archive/3f72b47466cc505b92c6e9b0ca2f5689e93a6a89.zip
The patch changes anonymous behavior so that:
- profile resolution can fall back from
api/v1/users/web_profile_info/toapi/v1/feed/user/{username}/username/; - anonymous post pagination uses
api/v1/feed/user/{id}/; - post pages use
max_idpagination; - items are converted through Instaloader's iPhone/mobile response model;
- failures from
web_profile_infocan reach the feed fallback immediately instead of waiting through a long retry/backoff loop.
No authenticated Instagram account is used.
Architecture
MCP Client
|
v
Vercel /api/mcp
|
v
Patched Instaloader (PR #2722)
|
v
Instagram mobile feed endpoint
MCP Tools
get_instagram_profile
Returns public metadata for an Instagram profile.
{
"username": "instagram"
}
get_instagram_posts
Returns recent public posts.
{
"username": "instagram",
"count": 3
}
debug_instagram_connection
Inspects the raw legacy web_profile_info HTTP response. This tool remains
available for diagnostics, but the patched Instaloader can fall back to the
mobile feed endpoint when that endpoint is refused.
{
"username": "instagram"
}
MCP Resources
instagram://profile/{username}
instagram://profile/{username}/posts
Cache
Default in-memory cache:
CACHE_TTL_SECONDS=600
DEFAULT_POST_COUNT=12
MAX_POST_COUNT=30
Set CACHE_TTL_SECONDS=0 to disable it.
Vercel instances are ephemeral, so this cache is per warm function instance, not a globally shared cache.
Deploy to Vercel
npm install -g vercel
vercel --prod
The included vercel.json deploys the function to São Paulo:
{
"$schema": "https://openapi.vercel.sh/vercel.json",
"regions": ["gru1"]
}
MCP endpoint:
https://YOUR-PROJECT.vercel.app/api/mcp
Health endpoint:
https://YOUR-PROJECT.vercel.app/api/health
Local development
python -m venv .venv
pip install -r requirements.txt
uvicorn api.index:app --reload
MCP endpoint:
http://127.0.0.1:8000/api/mcp
Cursor
.cursor/mcp.json:
{
"mcpServers": {
"instagram-public": {
"url": "https://YOUR-PROJECT.vercel.app/api/mcp"
}
}
}
Security
This repository contains no Instagram account credentials.
It does not contain or require:
- Instagram username/password credentials
sessionidcsrftokenfb_dtsg- authenticated account IDs
- saved Instaloader sessions
- Instagram cookies
Limitations
The server only queries public data and does not attempt to bypass private profiles.
Instagram may still rate-limit or block anonymous traffic based on the
deployment IP. The mobile-feed patch fixes the known broken anonymous
web_profile_info/timeline path, but it cannot guarantee that Instagram will
accept traffic from every hosting provider or IP range.
v2.1.1 profile compatibility fix
The anonymous mobile-feed fallback may omit optional profile fields that are
normally present in web_profile_info.
Version 2.1.1 reads profile properties defensively. Missing values such as
biography, external_url, or business-account metadata are returned as
null instead of causing get_instagram_profile to fail.
The post-fetching path is unchanged from v2.1.0.
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。