LinkedIn MCP for ChatGPT
Turn relationship context into a follow-up that continues the conversation.
LinkedIn MCP · 29 October 2026
A better follow-up starts before the first sentence
‘Just checking in’ is often a sign that the useful context has gone missing. What did the person ask for? Why did the conversation stop? Was there a specific commitment, or simply a suggestion to reconnect later?
A LinkedIn MCP workflow can help ChatGPT retrieve the relationship context available through a connected service before it drafts. Cold’s role is to bring the people, conversations and opportunities together; the writing should follow that evidence.
How ChatGPT connects to an MCP server
ChatGPT supports custom MCP servers as plugins with read and write tools. OpenAI’s current documentation describes adding the connection on the web, installing the resulting plugin and selecting it in a conversation. Workspace permissions and security restrictions apply.
Start from Cold’s MCP page for the options offered to your account. An assistant’s ability to connect to MCP does not mean that Cold is listed in its public directory or that every LinkedIn action is available. Check the tools exposed by your actual connection.
Recover the conversation before drafting
Begin with a named contact, a company and a time window. Ask ChatGPT to summarise the last substantive exchange and identify whether the next move belongs to you or the other person.
Try: ‘Use Cold to find my latest conversation with James at Ironwood Capital. What did we agree, what did I promise to send, and what has happened since? Show the supporting context. Do not draft a message yet.’
This separates research from writing. If the assistant cannot find the exchange, resolve that first rather than asking it to produce a polished message from incomplete information.
Draft one clear next step
Once the summary is right, ask: ‘Draft a concise LinkedIn follow-up that refers to the actual open item. Use my tone, include one clear next step and avoid implying that anything was agreed unless the conversation supports it. Show me the draft; do not send it.’
A good draft might reference a requested deck or propose the call you discussed. It should not invent urgency, interest or familiarity. The example contact is illustrative; use a real relationship and check the source material yourself.
Keep LinkedIn and email context together
Relationships frequently cross channels. An introduction may arrive on LinkedIn while the substantive exchange happens by email. If those sources are available through Cold, ask for a single timeline with the channel and date attached to each event.
That timeline can prevent an awkward follow-up after the person has already answered elsewhere. Ask ChatGPT to flag uncertainty when dates, people or conversation threads do not line up.
A draft is not a sent message
MCP is a connection standard, not a guarantee of messaging capability. The tools in your account determine whether a workflow can retrieve history, create drafts or take an action. Keep those steps explicit.
Review the recipient, wording and source context before sending anything. A useful first test is a read-only summary of a conversation you already know, followed by a draft you can edit yourself.
Put your next follow-up in context
Start with one relationship where there is a real reason to reconnect. Recover the history, choose the next step and then write. This makes the message an extension of the relationship rather than a generic outreach template.
Explore Cold MCP to see the context layer behind the workflow. The LinkedIn MCP Server overview also links to meeting preparation and research guides for other assistants.
Connection guidance
Platform guidance checked on 29 October 2026. Connection availability and permissions can vary by account.