LinkedIn MCP
Hosted LinkedIn MCP server
One remote MCP endpoint for LinkedIn. Add it to Claude Code, Cursor, Codex, ChatGPT or Claude, sign in, and your AI can read, search and answer your LinkedIn chats with your approval.
Remote MCP over Streamable HTTP with OAuth sign-in. Nothing to install or keep running.
https://api.connectchats.com/mcpHosted vs running your own LinkedIn MCP server
You can run an MCP server for LinkedIn yourself, or use a hosted one. Here is what changes.
ConnectChats (hosted)
- One URL. Nothing runs on your computer, so it keeps working when your laptop is closed.
- Sign in with OAuth from your MCP client. No tokens or session files to copy around.
- The same account and the same shared chats work in every MCP client you add it to.
- Sends are saved as drafts you approve, unless you turn on auto-send for a chat.
- Stored messages are encrypted with a key that only your passkey can unlock.
Running your own server
- You run the bridge and the server on your own machine or VPS, and keep both online.
- You handle the session, updates and reconnecting when LinkedIn drops the link.
- Each MCP client needs its own local setup, often a process started over stdio.
- You control the code and where the data lives.
- Approval, limits and prompt-injection safeguards are up to you.
The seven tools
Every ConnectChats MCP server exposes the same tools. Your AI picks the right one for what you ask.
list_linked_apps- Lists the chat apps you linked, such as LinkedIn, one line per account, with each account's connection state.
list_chats- Lists the chats you shared, unread first, with unread counts, the last message time and a preview.
read_messages- Reads messages in one shared chat: only what you have not read yet, or further back page by page.
search_messages- Searches message text across your shared chats, or within one chat app or one chat.
get_messages_to_summarize- Fetches the messages of a chat since you last read it, or since a time, so your AI can summarize them.
draft_reply- Saves a reply as a draft. Nothing is sent until you approve it in ConnectChats.
send_message- Sends right away only in chats with auto-send on. Anywhere else it saves a draft with a link to approve it.
Connect the LinkedIn MCP server to your client
Use the same server URL everywhere. The first time, your client opens a browser window so you can sign in.
Claude Code
Run this in your terminal, then type /mcp in Claude Code to sign in.
claude mcp add --transport http connectchats https://api.connectchats.com/mcpCursor
Add this to ~/.cursor/mcp.json, or .cursor/mcp.json in a project, then sign in when Cursor asks.
{
"mcpServers": {
"connectchats": {
"url": "https://api.connectchats.com/mcp"
}
}
}Codex
Run this in your terminal, then run codex mcp login connectchats to sign in.
codex mcp add connectchats --url https://api.connectchats.com/mcpApproval first, and messages treated as untrusted
-
Sends need your approval
send_message saves a draft that you approve in ConnectChats. It sends right away only in chats where you turned on auto-send, and you can turn that off at any time.
-
Messages are untrusted content
Every message the server returns is wrapped and marked as written by other people, so your AI treats text from LinkedIn as data, not as instructions to follow.
-
You choose what is shared
Your AI only sees the LinkedIn chats you share in the dashboard. Stored messages are encrypted with a key that only your passkey can unlock.
LinkedIn support
ConnectChats links to your LinkedIn messages using your signed in browser session, since LinkedIn offers no other way to share personal messages with another app. You choose which conversations your AI app can see.
- Your LinkedIn inbox is supported: one to one messages, group conversations and InMail you receive.
- Posts, comments and connection requests are not included, only messages.
- Each linked chat starts with its last 200 messages, then every new message arrives as it is sent.
- Sales Navigator and Recruiter inboxes are separate products and are not included.
Questions
How do I start using the LinkedIn MCP server?
Add https://api.connectchats.com/mcp to your MCP client, sign in, link LinkedIn from your dashboard and share the chats you want your AI to see.
What is a LinkedIn MCP server?
MCP (Model Context Protocol) is how AI apps call outside tools. A LinkedIn MCP server gives an AI app like Claude or Cursor tools to list your LinkedIn chats, read and search messages, and draft replies.
Do I need to run anything locally?
No. ConnectChats is a remote MCP server at https://api.connectchats.com/mcp. Your MCP client connects over HTTPS and signs you in with OAuth.
Which MCP clients work?
Any client that supports remote MCP servers over Streamable HTTP, including ChatGPT, Claude, Grok, Manus, and Muse.
Can the server send messages on its own?
No. send_message saves a draft you approve, and it sends directly only in chats where you turned on auto-send.
How does it handle prompt injection?
Messages come back wrapped as untrusted content, and every tool tells the model to treat them as data. Outside chats where you turned on auto-send, nothing is sent without your approval.
Is it free?
Yes. The free plan links two chat apps and works in every supported MCP client. No card needed.
Does it help me keep up with recruiter messages?
Yes. Your AI app can list the LinkedIn messages that still need an answer and draft polite replies you approve before they are sent.
Why does LinkedIn need browser cookies?
LinkedIn does not offer a sign-in that lets another app read your messages. Your browser session is how ConnectChats connects as you, and signing out of LinkedIn ends it.
Related pages
Use the LinkedIn MCP server in
More about LinkedIn
MCP servers for other chat apps
Stop switching apps. Ask your AI instead.
Add ConnectChats in one click. Your first two chat apps are free.
Get started free