LinkedIn Toolkit — Claude Code MCP server: what it does
MCP server that gives agents LinkedIn access through your own Chrome session, without headless fingerprints or bans.
claude mcp add --env LINKEDIN_TOOLKIT_HOME=<LINKEDIN_TOOLKIT_HOME> --transport stdio linkedin-toolkit -- npx -y linkedin-toolkit-mcpWhat LinkedIn Toolkit does
LinkedIn Toolkit routes agent calls through a Chrome extension into your logged-in LinkedIn session, using the same internal endpoints the page itself calls. The agent can search people, read profiles, send connection requests with notes, run message sequences, work the inbox, and save results to SQLite. Everything runs with human delays and hard daily caps: 100 invites, 150 messages, 500 profile visits, 1,000 search results.
Copilot mode is default: every write action lands in an approval queue in the extension popup; the agent sends nothing on its own. Caps live in the extension and no CLI flag or config can lift them. The extension unpacks to your local disk, data stays in chrome.storage.local, IndexedDB and a SQLite file on your machine.
Includes 39 MCP tools, 4 resources, 3 prompts, an HTTP API with OpenAPI 3.1, Node and Python clients, n8n nodes, 20 sequence templates and 6 skills.
What it helps with
- connect an agent to LinkedIn without headless browser bans
- pull search and Sales Navigator results into lists
- send connection requests with personalized notes through an approval queue
- process inbox threads with tagging and auto-replies
- export profiles and conversations to CSV or SQLite for CRM
Example prompts for the agent
- Search for all Head of Engineering in Berlin with Rust experience and save to a list called
berlin-rust-leads - Send connection requests to the first ten people in
berlin-rust-leadswith the note from templateoutreach-rust - Show unread inbox threads and tag anyone who mentioned the word hiring
How to install LinkedIn Toolkit
Platform CLI
claude mcp add --env LINKEDIN_TOOLKIT_HOME=<LINKEDIN_TOOLKIT_HOME> --transport stdio linkedin-toolkit -- npx -y linkedin-toolkit-mcpReplace <…> with your own values
Platform CLI
codex mcp add linkedin-toolkit --env LINKEDIN_TOOLKIT_HOME=<LINKEDIN_TOOLKIT_HOME> -- npx -y linkedin-toolkit-mcpReplace <…> with your own values
Config file
{
"mcpServers": {
"linkedin-toolkit": {
"command": "npx",
"args": [
"-y",
"linkedin-toolkit-mcp"
],
"env": {
"LINKEDIN_TOOLKIT_HOME": "<LINKEDIN_TOOLKIT_HOME>"
}
}
}
}Replace <…> with your own values
Config file
{
"servers": {
"linkedin-toolkit": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"linkedin-toolkit-mcp"
],
"env": {
"LINKEDIN_TOOLKIT_HOME": "<LINKEDIN_TOOLKIT_HOME>"
}
}
}
}Replace <…> with your own values
Platform CLI
gemini mcp add -e LINKEDIN_TOOLKIT_HOME=<LINKEDIN_TOOLKIT_HOME> linkedin-toolkit npx -y linkedin-toolkit-mcpReplace <…> with your own values
Step by step, with commands for each platform: How to add an MCP server to Claude Code, Codex, Cursor, VS Code, Gemini CLI and more
Questions about LinkedIn Toolkit
How do I install LinkedIn Toolkit in Claude Code?
Run claude mcp add linkedin-toolkit -- npx -y linkedin-toolkit-mcp, then open chrome://extensions, turn on Developer mode, click Load unpacked, and choose ~/.linkedin-toolkit/extension. Paste the token from the command output into the extension's Settings.
What does LinkedIn Toolkit do?
Gives agents LinkedIn access through your own Chrome session: search, profiles, connection requests with notes, messages, inbox, touch sequences, and export to SQLite. Everything with an approval queue and daily caps.
Why does LinkedIn Toolkit use a Chrome extension instead of a headless browser?
LinkedIn detects and bans headless browsers by fingerprints and datacenter IPs. The extension runs in your regular Chrome session, using the same internal endpoints the site does, at human pace and from your own IP.
