What is a LinkedIn MCP server?

How it works

The Model Context Protocol is an open standard for connecting AI assistants to outside tools and data. An MCP server publishes a list of named tools, each with a description and inputs. The assistant reads that list, picks a tool for your request and calls it with the permissions you granted when you signed in. On a LinkedIn MCP server, the tools act on an outreach account: its campaigns, prospects, connected LinkedIn and email accounts, inbox and reports.

Remote servers run over Streamable HTTP, so you add a URL to your assistant instead of installing software. LinkedCamp's server signs you in with OAuth 2.1 with PKCE (API key entered on LinkedCamp's authorization page) or an X-Api-Key / Bearer header.

What to watch out for

• Read or write: some servers only report. Check whether the tools can build and change campaigns, and which objects they cover.
• Approval: actions that reach people, such as launching a campaign or sending a message, should make the assistant ask you first.
• Scope across accounts: agencies should check whether one connection reaches every client, and whether the server keeps each client's data apart.
• Credentials: the assistant should never need your LinkedIn password. Connecting a LinkedIn account belongs in the tool's own app.

How LinkedCamp does it

LinkedCamp's server exposes 133 tools in 15 groups, from campaigns and sequences to the unified inbox, analytics, AI agents and sub-accounts: 64 read-only and 69 that change data, 28 of them destructive, so the assistant asks first. ChatGPT connects through a published plugin; Claude, Claude Code, Cursor, Windsurf, Zed and other clients connect to the same URL.

For lead gen agencies, an owner key lists every client sub-account, and 7 tools accept a sub-account id. The server rejects ids outside your own agency. Connecting a LinkedIn account, billing, team roles and stored credentials are never available through MCP.