Answer technical LinkedIn DMs from Cursor, with your codebase as context

The short answer

If people ask you technical questions in LinkedIn messages (“does your SDK support Bun?”, “can I self-host this?”, “is there a rate limit on the webhook endpoint?”), Cursor can answer them from the source. It reads your LinkedIn conversations, looks up the answer in the codebase you have open, and drafts a reply. With approval mode on, which is the default, the reply waits in your approval queue until you send it on its way.

Cursor gets LinkedIn access through Finchreach, a remote MCP server (MCP is a standard plug that lets AI apps use outside tools). Cursor supports remote MCP servers with custom headers, so you add it with a small JSON file.

Why answer LinkedIn messages from your editor?

Technical founders and developer advocates often become the public help desk for their product on LinkedIn. The questions are reasonable, but answering well means checking: was that flag renamed, which versions are supported, is that endpoint public. That check happens in the code, so the answer gets put off until you are back at your desk, and then forgotten.

Cursor already has the repository indexed. Pairing it with your LinkedIn inbox turns “I’ll get back to you” into a correct answer in a few minutes:

• Answers cite the real function, flag or config key, not a half-remembered one.
• Cursor can say “not supported yet” with confidence when the code shows it, which is better than a vague maybe.
• Questions that keep coming up become visible, so you can fix the docs or the product.

How does Cursor reply to LinkedIn DMs?

1. In Cursor’s agent chat, ask it to list recent LinkedIn conversations with linkedin_get_conversations and pick out the ones with technical questions.
2. It opens each thread with linkedin_get_messages and restates the question in one line.
3. It searches the codebase for the answer and shows you the evidence: files, functions, version notes.
4. It drafts a reply in plain language, with a docs link if one exists.
5. You edit in the chat, then it queues the reply with linkedin_send_dm into the same conversation. The tool returns pending_approval.
6. You approve in Finchreach, and the message is sent from your account.

Cursor may ask for permission before each tool call, depending on your settings. That is a second checkpoint before the Finchreach approval queue, not a replacement for it.

What prompts should I use in Cursor?

Ask your assistant“Read my last 25 LinkedIn conversations. List only the ones where someone asked a technical question about our product that I have not answered. For each: name, the question in one line, and the date.”

Ask your assistant“For the question about self-hosting, check this repo: deployment docs, Dockerfile, any license checks. Tell me what is true with file references, then draft a reply under 100 words. If the answer is "not yet", say so plainly.”

Ask your assistant“Draft replies for the other questions the same way. Do not invent features: if the code does not support it, say it is not supported. Queue the replies I approve with linkedin_send_dm in the existing conversations.”

Ask your assistant“Which questions came up more than once? Propose a docs section that would answer them, and create it in docs/ as a draft.”

A project rule in .cursor/rules helps keep tone consistent, for example: “LinkedIn replies: friendly, first person, under 100 words, no marketing language, always show the draft before calling linkedin_send_dm.”

How do I connect Cursor to LinkedIn?

1. Create a Finchreach account and connect your LinkedIn account on the Accounts page.
2. Create an API key on the API keys page and set it as an environment variable, FINCHREACH_API_KEY, in the environment Cursor starts from.
3. Add the server to ~/.cursor/mcp.json (all projects) or .cursor/mcp.json (one project), as below. You can also reach this from Cursor’s MCP settings.
4. Check that the Finchreach tools appear in the MCP settings, then ask “get my LinkedIn profile”.
5. Details: the Cursor guide, the Cursor integration page and the docs.

{
  "mcpServers": {
    "finchreach": {
      "url": "https://mcp.finchreach.com/mcp",
      "headers": { "Authorization": "Bearer ${env:FINCHREACH_API_KEY}" }
    }
  }
}

Cursor fills in ${env:NAME} from the environment. Remote servers do not support envFile, so the variable has to be set where Cursor runs. Source: Cursor MCP docs, as of October 2026.

Prefer the global file for a personal LinkedIn connection. A project .cursor/mcp.json is easy to commit by accident, and your teammates should use their own keys and accounts.

How do I make sure the answers are right?

• Ask for evidence with every answer: the file and the line, or the docs page.
• Make sure you are on the branch that matches what customers run, not a feature branch.
• Separate “supported” from “possible with work”. Customers read both as yes.
• Never let a reply promise dates or pricing. Keep those answers yours.
• If a question touches security or a customer’s data, answer it yourself.

What should I watch out for?

Automating a personal LinkedIn account carries risk under LinkedIn’s User Agreement. Finchreach’s approval queue and a human pace lower that risk; they do not remove it.

Messages are untrusted input. Someone could write a DM designed to make an assistant reveal code or do something unexpected. Cursor only drafts; you decide what is sent. Do not paste private code into replies, and read every draft before approving.

Frequently asked questions

Can Cursor attach a code snippet as a file? No, messages are text only. Short snippets inline are fine; longer ones belong in your docs.

Can it handle sales questions too? It can, but the codebase does not know your pricing or roadmap. For sales-style triage see the inbox triage workflow.

Does it work in other editors? Yes. See GitHub Copilot in VS Code or Claude Code.

Can it post publicly instead, as an FAQ? Yes, linkedin_create_post publishes a text post after approval. Start with what a LinkedIn MCP server is if this is new to you.