linkedin-mcp
Enables LLMs to manage LinkedIn posts and view analytics through a set of MCP tools, including creating, reading, and deleting posts, checking authentication, and retrieving profile info.
README
LinkedIn MCP Server
MCP server for LinkedIn post management and analytics. Deploy on Vercel, connect to Claude or any MCP-compatible LLM.
Architecture
Claude / LLM
│
│ POST /mcp (JSON-RPC)
▼
LinkedIn MCP Server (Vercel)
│
│ OAuth 2.0 + Bearer token
▼
LinkedIn API v2
Step-by-Step Setup
STEP 1 — Create LinkedIn App
- Go to https://developer.linkedin.com/apps
- Click Create App
- Fill in: App Name, LinkedIn Page (your company), App Logo
- Click Create App
- Go to Auth tab → note your Client ID and Client Secret
- Under OAuth 2.0 settings → Authorized redirect URLs, add:
- Local dev:
http://localhost:3000/auth/callback - Production:
https://your-app.vercel.app/auth/callback
- Local dev:
- Go to Products tab → request access to:
- Sign In with LinkedIn using OpenID Connect (instant approval)
- Share on LinkedIn → gives you
w_member_socialscope (instant) - Marketing Developer Platform → only if you need impression analytics (requires manual review)
STEP 2 — Clone and Install
git clone <your-repo>
cd linkedin-mcp-server
npm install
STEP 3 — Configure Environment
cp .env.example .env
Edit .env:
LINKEDIN_CLIENT_ID=86xxxxxxxxxx
LINKEDIN_CLIENT_SECRET=WPxxxxxxxxxx
LINKEDIN_REDIRECT_URI=http://localhost:3000/auth/callback
PORT=3000
STEP 4 — Build and Run Locally
npm run build
npm start
Server starts at http://localhost:3000
Endpoints:
GET /health— health checkGET /auth/linkedin?user_id=YOUR_ID— start OAuth flowGET /auth/callback— LinkedIn redirects here automaticallyPOST /mcp— MCP endpoint for LLM tool calls
STEP 5 — Authenticate a LinkedIn Account
Open in browser:
http://localhost:3000/auth/linkedin?user_id=herbert
This redirects you to LinkedIn → you log in → LinkedIn redirects to your callback → token is stored.
Response you'll see:
{
"success": true,
"message": "LinkedIn authenticated for user herbert",
"person_id": "abc123xyz",
"expires_in_days": 60
}
Save the person_id — you'll need it when calling post tools.
STEP 6 — Deploy to Vercel
npm install -g vercel
vercel login
# Set secrets
vercel env add LINKEDIN_CLIENT_ID
vercel env add LINKEDIN_CLIENT_SECRET
vercel env add LINKEDIN_REDIRECT_URI # set to https://your-app.vercel.app/auth/callback
# Deploy
npm run build
vercel --prod
After deploy, update your LinkedIn app's redirect URI to the production URL.
STEP 7 — Connect to Claude
In Claude.ai → Settings → Connectors → Add MCP Server:
URL: https://your-app.vercel.app/mcp
That's your MCP URL.
Available MCP Tools
| Tool | What it does |
|---|---|
linkedin_check_auth |
Check if a user is authenticated |
linkedin_get_profile |
Get profile info + LinkedIn person ID |
linkedin_create_post |
Publish a post to LinkedIn |
linkedin_get_my_posts |
Fetch your recent posts |
linkedin_get_post_analytics |
Get likes/comments on a post |
linkedin_delete_post |
Delete a post you own |
Example Claude Usage
Once connected, you can say:
"Check if user herbert is authenticated on LinkedIn"
"Post to LinkedIn as herbert: 'Excited to share our latest Oracle HCM implementation update...'"
"Get the last 5 LinkedIn posts for herbert and show me engagement"
Production: Replace In-Memory Token Store with Supabase
The current code stores tokens in memory (lost on restart). For production:
// src/services/linkedin.ts — replace storeToken/getToken with:
import { createClient } from '@supabase/supabase-js';
const supabase = createClient(
process.env.SUPABASE_URL!,
process.env.SUPABASE_SERVICE_KEY!
);
export async function storeToken(userId: string, token: LinkedInToken) {
await supabase.from('linkedin_tokens').upsert({
user_id: userId,
access_token: token.access_token,
refresh_token: token.refresh_token,
expires_at: new Date(token.expires_at).toISOString(),
person_id: token.person_id,
});
}
export async function getToken(userId: string): Promise<LinkedInToken | undefined> {
const { data } = await supabase
.from('linkedin_tokens')
.select('*')
.eq('user_id', userId)
.single();
if (!data) return undefined;
return {
access_token: data.access_token,
refresh_token: data.refresh_token,
expires_at: new Date(data.expires_at).getTime(),
person_id: data.person_id,
};
}
Supabase table:
create table linkedin_tokens (
user_id text primary key,
access_token text not null,
refresh_token text,
expires_at timestamptz not null,
person_id text,
created_at timestamptz default now(),
updated_at timestamptz default now()
);
LinkedIn API Scope Reference
| Scope | What it unlocks | Approval |
|---|---|---|
r_liteprofile |
Name, ID, photo | Instant |
r_emailaddress |
Email address | Instant |
w_member_social |
Post as yourself | Instant |
r_organization_social |
Read company page posts | Manual review |
rw_organization_admin |
Post as company page | Manual review |
r_ads_reporting |
Impression/click analytics | Marketing partner only |
Token Expiry
- Access tokens expire in 60 days
- Refresh tokens expire in 365 days (if
offline_accessgranted) - The server auto-refreshes tokens before they expire
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。