@crelora/onelin-bridge

OneLin Bridge (MCP Server)

Bridge server for the OneLin LinkedIn Extension. This server allows AI agents (via the Model Context Protocol) to interact with a user's LinkedIn messaging tabs in real-time.

Protocol Contract

The bridge manages communication between an MCP Client (e.g. OpenClaw) and the Chrome Extension via a local WebSocket.

MCP Tools (Agent Interface)

Tool	Parameters	Description
get_status	None	Checks if the extension is connected and lists open LinkedIn tabs.
get_conversation_context	tabId?	Retrieves conversation messages and prospect profile info from the current chat.
get_draft	tabId?	Reads the current text in the LinkedIn message input field.
write_reply	text, tabId?	Injects text into the message editor.
send_message	tabId?	Clicks the LinkedIn 'Send' button.
generate_reply	modeType?, modeName?, customInstruction?, tabId?	Delegate to the extension's built-in specialized LLM generation logic.
list_linkedin_tabs	None	Lists all open LinkedIn tabs (id, url, title).

Setup for Agents (e.g. Claude Desktop, OpenClaw)

Once this package is published to npm, it can be added to your client configuration:

{
  "mcpServers": {
    "onelin": {
      "command": "npx",
      "args": ["-y", "@crelora/onelin-bridge"]
    }
  }
}

How it works

• The Bridge (this server) runs locally and listens on ws://localhost:3003.
• The Extension (via its offscreen doc) connects to the bridge.
• The Agent calls an MCP tool on the bridge.
• The Bridge serializes the command and sends it over the WebSocket to the extension.
• The Extension executes DOM actions on LinkedIn and returns the data back to the bridge.

Security

• The WebSocket server only binds to localhost.
• Commands only affect tabs with linkedin.com URLs.
• No LinkedIn credentials/cookies ever leave the browser.