Linked API MCP server is an MCP (Model Context Protocol) server whose purpose is to connect your LinkedIn account to AI assistants such as Claude, Cursor, and VS Code, enabling the AI to control the LinkedIn account and retrieve real-time data. The project is published on GitHub by Linked-API and is licensed under MIT.
Purpose The server targets scenarios that require automating LinkedIn operations. The official typical use cases include:
- Sales automation assistant: lets the AI search for potential clients, view their profiles, and draft personalized outreach messages. For example, search for "software engineers at companies in San Francisco with 50-200 employees," analyze their backgrounds, and suggest reasonable connection messages.
- Recruiting assistant: lets the assistant search for candidates with specific skills, view their experience, and send initial outreach.
- Conversation assistant: the AI can read your existing LinkedIn conversations, understand the context, suggest natural replies, and even send follow-up messages.
- Market research assistant: collect data on companies, employees, and recent activity for competitive analysis and industry trend insights.
How to use it Officially, it is recommended to first read three guides:
- Installation: set up MCP in Claude, Cursor, VS Code, or Windsurf.
- Available tools: learn about all the LinkedIn tools the assistant can call.
- Usage examples: view real examples to get started quickly.
Long-running operations Linked API operations are executed through a cloud browser and are queued like ordinary automation tasks. Many operations take several minutes, especially searches and profile retrievals with optional data. If a tool returns status, workflowId, operationName, and message, it means the operation is still running. In this case, do not retry the original tool, otherwise duplicate tasks may be queued. Instead, call get_workflow_result with the exact workflowId and operationName until the final result is returned. By default, get_workflow_result performs long polling until the workflow completes or the current MCP client's request budget is exhausted; passing waitSeconds: 0 retrieves a single snapshot immediately.Retrieve a single snapshot immediately.
Limitations Operations rely on a cloud browser and are queued for execution, so they may take a long time; some operations require polling to retrieve results and should not be submitted repeatedly. Other features and limitation information are not yet provided.
