linkedin-agent

Read your LinkedIn profile and publish posts to your feed, through chat. Pick which AI model does the thinking — OpenAI, Claude, or DeepSeek.

your prompt
               │
               ▼
┌─────────────────────────────┐
│       linkedin-agent        │   reads your prompt
└──────────────┬──────────────┘
               │
               ▼
┌─────────────────────────────┐
│           bridge            │   openai / claude / deepseek — talks to the AI model
└──────────────┬──────────────┘
               │
               ▼
┌─────────────────────────────┐
│     linkedin-mcp-server     │   reads/posts your profile
└──────────────┬──────────────┘
               │
               ▼
an answer, or a new post on your feed

What you need

• Node.js 18+
• A LinkedIn Developer app (free, a few minutes — see step 1 below)
• An API key for at least one of: OpenAI, Anthropic (Claude), DeepSeek

Setup

1. Create a LinkedIn app:

1. Go to linkedin.com/developers/apps → Create app.
2. Under Products, add "Sign In with LinkedIn using OpenID Connect" and "Share on LinkedIn".
3. Under Auth, add an Authorized redirect URL: http://localhost:3300/callback.
4. Copy the app's Client ID and Client Secret.

2. Authorize it (one time — run from a folder you'll reuse in step 3):

LINKEDIN_CLIENT_ID=<id> LINKEDIN_CLIENT_SECRET=<secret> \
  npx -p @clickonsearch/linkedin-mcp-server linkedin-authorize

This opens a link — sign in, approve, done. It saves a token to linkedin-token.json in the current folder, so step 3 needs to run from this same folder.

3. Start the LinkedIn connection (own terminal, keep it running):

LINKEDIN_CLIENT_ID=<id> LINKEDIN_CLIENT_SECRET=<secret> \
  npx @clickonsearch/linkedin-mcp-server

Use it: command line

ANTHROPIC_API_KEY=<your-key> npx @clickonsearch/linkedin-agent --provider claude "what does my profile look like?"

Swap in OPENAI_API_KEY + --provider openai, or DEEPSEEK_API_KEY + --provider deepseek, to use a different model.

Use it: browser chat

ANTHROPIC_API_KEY=<your-key> npx -p @clickonsearch/linkedin-agent linkedin-agent-serve

Then open http://localhost:3200 and start chatting.

Use it: embedded in another tool

The browser chat server speaks the AG-UI protocol (POST /agent, streaming SSE) — any AG-UI-compatible client can drive it as a backend, not just the bundled page.

Config reference

| Env var | Required | Default | | -------------------- | -------------------- | -------------------------- | | PROVIDER | no | openai | | OPENAI_API_KEY | if using OpenAI | | | OPENAI_MODEL | no | gpt-4.1 | | ANTHROPIC_API_KEY | if using Claude | | | ANTHROPIC_MODEL | no | claude-sonnet-5 | | DEEPSEEK_API_KEY | if using DeepSeek | | | DEEPSEEK_MODEL | no | deepseek-chat | | LINKEDIN_MCP_URL | no | http://127.0.0.1:4300 | | LINKEDIN_MCP_TOKEN | only if the server needs one | | | PORT | no, browser chat only | 3200 |

Good to know

This agent can only do two things: read your own profile, and post to your own feed. It can't send LinkedIn messages (no public API exists for that, for anyone) or reply to other people's posts (needs special LinkedIn approval most apps don't have). If you ask it to do either, it'll tell you it can't rather than pretend.

For developers

Running from source:

# terminal 1
cd mcp-servers/linkedin-mcp-server
npm install
cp .env.example .env   # fill in LINKEDIN_CLIENT_ID / LINKEDIN_CLIENT_SECRET
npm run dev:authorize   # opens a consent link in your browser, one time
npm run dev

# terminal 2
cd agents/linkedin-agent
npm install
cp .env.example .env    # fill in a provider key
npm run dev -- --provider claude "what does my profile look like?"
npm run serve             # or browser chat

Same shape as github-agent: both the CLI and the browser chat UI (an AG-UI protocol server, same POST /agent shape) call through to linkedin-mcp-server's two tools.