linkedin-mcp
MCP server for LinkedIn, enabling posting, liking, commenting, and profile retrieval via official OAuth 2.0. Features CLI, HTTP admin, and optional Telegram bot.
README
linkedin-mcp
0 Trust · 100% Control | 0 Magic · 100% Transparency | 0 Hardcoding · 100% Flexibility
A Model Context Protocol (MCP) server for LinkedIn — official OAuth 2.0, intent-first tool surface, three-level admin (CLI · HTTP · Telegram). Designed for the KpihX homelab stack; deployment-ready with Docker + Traefik.
Why this exists
LinkedIn has no general-purpose developer API for personal accounts. The official OAuth surface (openid profile email w_member_social) covers everything an individual power-user needs:
- Post, delete, like, comment on content
- Read own profile and auth status
- Guide agents through the flow with
linkedin_guide
This MCP wraps that surface with full operational visibility: a CLI admin, an HTTP admin API, and an optional Telegram bot — the same architecture as whats-mcp, tick-mcp, mail-mcp.
Architecture
linkedin-mcp/
├── src/
│ ├── main.js ← entry: serve (stdio) | serve-http
│ ├── admin.js ← entry: CLI admin (linkedin-admin)
│ ├── config.js ← loadConfig() — deep merge defaults → config.json → env
│ ├── token.js ← save / load / clear / summary (~/.mcps/linkedin/token.json)
│ ├── client.js ← LinkedInClient — official REST API calls
│ ├── server.js ← createMcpServer() — MCP SDK wiring
│ ├── http_app.js ← Express HTTP app: /health /admin/* /mcp
│ ├── .env.example ← required secrets template
│ └── admin/
│ ├── cli.js ← Commander CLI (auth / status / logout / logs / health / urls
│ │ / client-id / client-secret / token)
│ ├── service.js ← unified admin kernel — single source of truth for all surfaces
│ ├── oauth.js ← LinkedIn OAuth 2.0 + OIDC flow (configurable callback port)
│ └── telegram.js ← optional Telegram bot admin (15 commands)
├── tools/
│ ├── registry.js ← listTools() + callTool()
│ ├── guide.js ← linkedin_guide
│ ├── profile.js ← linkedin_get_profile, linkedin_auth_status
│ ├── posts.js ← linkedin_create_post, linkedin_create_image_post, linkedin_delete_post
│ └── social.js ← linkedin_like_post, linkedin_create_comment
├── tests/ ← bun test (61 tests, 0 deps on live API)
├── deploy/
│ ├── docker-compose.yml
│ └── docker-compose.override.example.yml
├── Dockerfile
├── .gitlab-ci.yml
└── config.json ← non-secret defaults (port 8095, OAuth port 3001, state dir, scopes)
Transport strategy:
Agent / Claude / Gemini / Codex / Copilot / Vibe
│
├─── HTTP (homelab) ──→ https://linkedin.kpihx-labs.com/mcp (primary)
└─── stdio (local) ──→ ~/.bun/bin/linkedin-mcp serve (fallback)
Tools (8)
| Tool | Description |
|---|---|
linkedin_guide |
Orientation — all tools, auth flow, quick start |
linkedin_auth_status |
Token presence, validity, days remaining, profile |
linkedin_get_profile |
Fetch own LinkedIn profile (name, email, sub) |
linkedin_create_post |
Create a text post (PUBLIC or CONNECTIONS) |
linkedin_create_image_post |
Create a post with an attached image |
linkedin_delete_post |
Delete a post by URN |
linkedin_like_post |
Like a post by URN |
linkedin_create_comment |
Add a comment to a post by URN |
Admin interfaces
CLI (linkedin-admin)
# Authentication
linkedin-admin auth [--port N] # OAuth flow — opens browser, saves token (default port 3001)
linkedin-admin token set <value> # Set access token directly (fetches profile, no browser)
linkedin-admin token unset # Clear stored token (same as logout)
linkedin-admin logout # Clear stored token
# Credential management (stored in admin env file)
linkedin-admin client-id set <v> # Set LINKEDIN_CLIENT_ID
linkedin-admin client-id unset # Clear LINKEDIN_CLIENT_ID
linkedin-admin client-secret set <v>
linkedin-admin client-secret unset
# Status & observability
linkedin-admin status # Credentials table + token summary
linkedin-admin status --json # Machine-readable JSON
linkedin-admin logs [--lines 50] # Tail admin log
linkedin-admin health # HTTP server reachability
linkedin-admin urls # All public/private endpoints
HTTP admin
| Endpoint | Method | Description |
|---|---|---|
/health |
GET | Liveness probe |
/admin/status |
GET | Full runtime status (token + Telegram state) |
/admin/help |
GET | All commands reference |
/admin/logs?lines=50 |
GET | Recent admin log tail |
/admin/client-id/set |
POST | Body: { "value": "..." } |
/admin/client-id/unset |
POST | Clear LINKEDIN_CLIENT_ID |
/admin/client-secret/set |
POST | Body: { "value": "..." } |
/admin/client-secret/unset |
POST | Clear LINKEDIN_CLIENT_SECRET |
/admin/token/set |
POST | Body: { "value": "..." } — sets token + fetches profile |
/admin/token/unset |
POST | Clear stored token |
/mcp |
POST/GET/DELETE | MCP Streamable HTTP transport |
Telegram bot (optional)
Set TELEGRAM_LINKEDIN_HOMELAB_TOKEN + TELEGRAM_CHAT_IDS — server polls every 5 s.
| Command | Action |
|---|---|
/start /help |
Full command reference |
/status |
Credentials + token status |
/health |
HTTP server health |
/urls |
Endpoint map |
/logs [n] |
Last n admin log lines (default 20) |
/token_set <value> |
Set access token (fetches profile) |
/token_unset |
Clear stored token |
/client_id_set <value> |
Set LINKEDIN_CLIENT_ID |
/client_id_unset |
Clear LINKEDIN_CLIENT_ID |
/client_secret_set <value> |
Set LINKEDIN_CLIENT_SECRET |
/client_secret_unset |
Clear LINKEDIN_CLIENT_SECRET |
Quick start
1. LinkedIn App setup
- Go to LinkedIn Developer Portal
- Create an app → add products: "Sign In with LinkedIn using OpenID Connect" + "Share on LinkedIn"
- Set redirect URI:
http://localhost:3001/callback - Note
Client IDandClient Secret
2. Configure secrets
# Option A — via admin CLI (written to admin env file)
linkedin-admin client-id set <your_client_id>
linkedin-admin client-secret set <your_client_secret>
# Option B — via .env file
cp src/.env.example src/.env
# Fill in LINKEDIN_CLIENT_ID and LINKEDIN_CLIENT_SECRET
# Option C — via bw-env / kshrc (injected at login shell)
3. Install & authenticate
bun install
bun link # editable install: ~/.bun/bin/linkedin-mcp + linkedin-admin
# Full OAuth flow (browser consent)
linkedin-admin auth # default port 3001
linkedin-admin auth --port 3002 # if 3001 is busy
# Or set access token directly (server/headless contexts)
linkedin-admin token set <your_access_token>
linkedin-admin status # verify credentials and token
4. Start MCP server
# stdio (for agent config)
linkedin-mcp serve
# HTTP (for homelab deployment)
linkedin-mcp serve-http
Docker / Homelab deployment
cd deploy
cp docker-compose.override.example.yml docker-compose.override.yml
# Edit override with your secrets (or rely on CI env injection)
docker compose up -d
Traefik labels in deploy/docker-compose.yml expose:
https://linkedin.kpihx-labs.com(primary private trusted route)https://linkedin.homelab(fallback)
The container persists all state in /data (Docker volume):
/data/state/token.json— OAuth token/data/linkedin-admin.env— credentials written bytoken set/client-id set
Agent registration
Claude Code (~/.claude.json)
"linkedin-mcp": {
"type": "http",
"url": "https://linkedin.kpihx-labs.com/mcp"
},
"linkedin-mcp--fallback": {
"command": "/home/kpihx/.bun/bin/linkedin-mcp",
"args": ["serve"]
}
Gemini (~/.gemini/extensions/linkedin_mcp/gemini-extension.json)
{
"name": "linkedin-mcp",
"version": "0.2.1",
"mcpServers": {
"linkedin-mcp": { "httpUrl": "https://linkedin.kpihx-labs.com/mcp" },
"linkedin-mcp--fallback": {
"command": "/home/kpihx/.bun/bin/linkedin-mcp",
"args": ["serve"]
}
}
}
Codex (~/.codex/config.toml)
[mcp_servers.linkedin_mcp]
url = "https://linkedin.kpihx-labs.com/mcp"
[mcp_servers.linkedin_mcp_fallback]
command = "/home/kpihx/.bun/bin/linkedin-mcp"
args = ["serve"]
Copilot (~/.copilot/mcp-config.json)
"linkedin_mcp": { "type": "http", "url": "https://linkedin.kpihx-labs.com/mcp" },
"linkedin_mcp_fallback": {
"type": "stdio",
"command": "/home/kpihx/.bun/bin/linkedin-mcp",
"args": ["serve"]
}
Vibe (~/.vibe/config.toml)
[[mcp_servers]]
name = "linkedin"
transport = "http"
url = "https://linkedin.kpihx-labs.com/mcp"
[[mcp_servers]]
name = "linkedin_fallback"
transport = "stdio"
command = "/home/kpihx/.bun/bin/linkedin-mcp"
args = ["serve"]
Tests
bun test # 61 tests, 5 files, 0 live API calls
bun test --watch # watch mode
All tests use mocked fetch and temp directories — no credentials needed.
Token lifecycle
LinkedIn personal OAuth tokens expire in 60 days. There is no refresh token for non-Partner apps.
Day 0 → linkedin-admin auth (OAuth flow, browser consent)
OR linkedin-admin token set (direct token, headless/server)
Day 55+ → linkedin-admin status (shows days_left)
Day 60 → token invalid, re-run auth
Both flows produce a complete token.json with member_id populated — required for all posting tools.
License
MIT — see LICENSE.
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.