How do you set up a LinkedIn MCP server with Claude?
LinkedBoost editorial Reviewed
Add LinkedBoost's remote endpoint, https://mcp.linkedboost.ai/mcp, through Claude's custom connector flow, authorize the workspace, and connect your LinkedIn account through the provided setup path. Then verify the intended workspace and account using actual tool results. A connector listed in settings or an assistant returning plausible counts is not sufficient evidence.
1. Add the remote connector
In Claude, open Customize, then Connectors, and use Add custom connector. Enter https://mcp.linkedboost.ai/mcp. Organization-managed accounts may require an owner to add the connector before members can connect.
Follow Anthropic's current guide for the interface and account you use. Remote connector configuration is distinct from running a local MCP process or configuring Claude Code.
2. Authorize the intended workspace
Complete the sign-in and authorization flow opened by the connector. Confirm you are using the workspace intended for this task. Do not paste passwords, session cookies or tokens into the conversation.
Enable the connector for the conversation if needed. Ask the assistant to use the workspace overview tool and inspect the returned account or workspace identity, not only its summary.
3. Link LinkedIn separately
Ask to connect your LinkedIn account and follow the link provided by the account-connection tool. Complete any requested steps in that flow. Workspace authorization and LinkedIn connection are separate states.
Run a connection-status check and confirm the correct person and account. A valid empty workspace can return zero counts. A nonzero number is not proof of connection unless it comes from the actual intended tool result.
4. Verify a bounded workflow
Read a lead group or existing conversation you are authorized to inspect. Compare the response with the account and record you intended. Prepare a draft without authorizing a send or enrollment yet.
Review the target, message, history and operation. Creating a sequence, enrolling people and sending a message are distinct actions. Before authorizing enrollment, inspect which later steps it schedules and how they can be paused.
Troubleshoot the state that actually failed
If the connector is unavailable, check whether it is enabled and allowed for the organization. If a workspace call is unauthorized, repair workspace authorization. If LinkedIn is disconnected, follow the account reconnection path.
LinkedBoost works with Claude, ChatGPT, Instinct and Meta Muse. Connection setup and available controls differ by assistant.
This page covers Claude's remote connector. Do not reuse its UI steps as instructions for every other assistant. LinkedIn account restrictions remain a separate issue from connector setup.
Connection states and evidence
| State | Evidence to inspect |
|---|---|
| Connector configured | Correct remote URL and enabled connector |
| Workspace authorized | A tool read from the intended workspace |
| LinkedIn linked | Account identity and current connection status |
| Workflow ready | Correct record, draft, history and authorization boundary |
| Action complete | Completed result for the exact operation, not only a queue entry |
Frequently asked questions
What is the LinkedBoost MCP endpoint?
https://mcp.linkedboost.ai/mcp is the hosted endpoint documented for the connector.
Does adding the connector connect LinkedIn?
No. Workspace authorization and linking the LinkedIn account are separate steps.
Do zero counts mean the connection failed?
No. An empty workspace can return zero. Check the actual tool result, workspace identity and account status.
Can I use ChatGPT instead?
LinkedBoost works with Claude, ChatGPT, Instinct and Meta Muse. Connection setup and available controls differ by assistant.
Sources and references
Open the setup guide
Follow the connector setup, then verify the workspace and LinkedIn account before preparing an action.
Related answers
- What can an AI agent actually do on LinkedIn?Map LinkedBoost's research, sequence, message, inbox and activity tools to practical workflows. Check approval and completion for each action.
- Which LinkedIn MCP server should I use with Claude?Compare LinkedIn MCP servers by the action you need, how they connect to LinkedIn, and what you must verify before using one with Claude.
- What breaks when you run MCP servers in production?Plan for ambiguous tools, expired credentials, persistent state and uncertain writes. Use observable results and bounded retries when operating MCP workflows.