# How to automate LinkedIn lead search with AI
> Connect an LLM to live LinkedIn search through an API or MCP server, score profiles against your customer profile, and send relevant outreach.
- HTML: https://getkinlead.com/manage-linkedin-outbound-with-api
- Markdown: https://getkinlead.com/manage-linkedin-outbound-with-api.md

**Short answer:** Automating LinkedIn lead search means giving an AI model a tool that can run LinkedIn searches and read profiles, then setting a scoring rule that keeps only the people who match your customer profile. The AI can draft outreach from the reason each lead was found, and you can keep the work inside limits you set. You still need to check the list before messages go out.

## Define the customer you want before you search

An AI will search for whatever you describe. If the description is loose, you get a list of people who share a job title and nothing else. Start with the accounts you already close and write down the details that repeat: company size, market, role, and the event that makes them likely to buy now. A company that just hired a sales leader is a different lead from a company that has never hired one.

Put the description in plain language. Include the words people use about themselves, not internal categories. 'Founder of a B2B software company with 10 to 50 people' is easier to check than 'SMB SaaS leader.' Add one or two buying signals you can see on LinkedIn, such as a post about a problem, a job ad for a role that reports to the buyer, or a recent funding announcement.

**Weak customer description**

- Founders in SaaS
- Marketing managers in the US
- B2B companies

**Useful customer description**

- Founder of a B2B software company with 10 to 50 employees in the UK who posted about hiring a first salesperson
- VP of demand generation at a fintech with 200 to 500 employees that uses HubSpot
- Owner of a 5 to 10 person agency in New York who writes about AI automation

## Connect the AI to live LinkedIn data

An LLM cannot open LinkedIn and run a search on its own. It needs a tool that can run the search and return profiles. The tool can be a REST API or an MCP server. The model calls a function such as search_leads or get_profile, and the tool returns structured data the model can read.

Use a model that supports tool use. Claude, ChatGPT, Cursor and Claude Code can all call MCP tools. The setup is usually a server address, an API key, and a permission step. Test with a small query before you run a large search.

1. **Pick a model that can call tools.** Use a model with function calling or tool use. The model does not need to browse LinkedIn itself. It needs to call your search tool and read the result.
2. **Connect the API or MCP server.** Add the server endpoint and API key in your client. For chat apps, use OAuth. For scripts, use an API key. Keep the key out of shared documents.
3. **Run a small search.** Ask for three people who match a narrow description. Check the profiles yourself. If the results are wrong, fix the description before you add scoring.
4. **Check the data you get back.** You want the profile URL, headline, company, location, and the reason the person matched. If the tool returns only names, you cannot score or write a relevant message.

## Set a scoring rule that rejects weak leads

Scoring turns a search into a list you can use. You give the model a rule and a fit line. The rule should reward the attributes that matter and remove the people who fail a hard requirement. A lead below the fit line is not kept.

Keep the rule short enough to explain to a teammate. If you cannot say why a profile scored 80, the rule is too vague. Start with the must-haves, then add the signals. Review the first twenty scores by hand and adjust the rule before you trust it.

> Keep the fit line high enough that you would still send the message to every lead above it.

```
System prompt: You score LinkedIn profiles for a B2B software company.

Return a score from 0 to 100 and one sentence that explains the score.

Add points:
+40 if the person is a founder, owner, or head of sales.
+30 if the company sells B2B software.
+20 if the company has 10 to 50 employees.
+10 if the person posted about hiring, pipeline, or outbound in the last 90 days.

Remove the lead:
If the company is a consumer brand, return 0.
If the person is a student or a vendor, return 0.
If the location is outside the UK or EU, return 0.

Do not guess. If a field is missing, say so and score lower.
```

## Review the leads before any message goes out

Automation is useful for finding and sorting. It is not a reason to skip the review. Open the leads that scored above your line and check the current role and company fit, then confirm the reason the lead was found is real. A profile that says 'founder' but lists a closed company is not a lead.

Put the leads into groups by source or campaign. A group from a competitor's post comments needs different language from a group from a job ad. Keep the post URL and the comment with the lead. When you write the message, you can use the exact context instead of a generic compliment.

| | Check | What to look for | What to do |
| --- | --- | --- | --- |
| Current role | The person lists the role and company on their profile | Skip if the role changed or the company is gone |
| Company fit | The company matches the size, market, and location in your description | Move to a lower group or remove |
| Reason for the lead | A post, comment, job ad, or profile detail that you can name | Keep the URL and the detail with the lead |
| Contact history | No open invitation, no recent message, no reply from the same person | Wait or choose another lead |

## Write outreach that starts with the reason

The first line should tell the person why you are in their inbox. If they commented on a competitor's post, say that. If they posted a job ad for an SDR, say that. The reason is the proof that you did not paste the same message to a thousand people.

Keep the message short. Ask for a small next step. Do not explain your whole product in the first message. The goal is a reply. A demo comes later. If the person replies, answer the question they asked before you pitch.

- **draft:** Hi Sarah, I saw your comment on Northwind's post about hiring SDRs in London. You said the hard part is finding people who can write a real email.
- **draft:** I work on outbound systems for teams your size. Would it help if I sent over the checklist we use before a first SDR hire?
- **them:** Yes, send it. We are hiring now and the last two hires did not work out.
- **draft:** I will send it today. Is the role reporting to you or to the sales lead?

> Do not write a first message that could go to anyone. If you cannot name the reason in one line, the lead is not ready.

## Doing this with Kinlead

Kinlead finds people on LinkedIn who match the customer you describe or your website, then scores each lead from 0 to 100 against that profile. Leads below the fit line are not kept. Leads are kept in lead groups, and any campaign can point at any group. You can export leads to CSV or import a CSV of LinkedIn profile URLs.

Outreach runs from your own LinkedIn account through Unipile, a third-party LinkedIn provider. Kinlead never stores your LinkedIn password. It sends at a human pace, with at most one action every 5 minutes, inside the daily limits and send window you set. You can use the REST Agent API or MCP server to find leads, check replies and draft answers from Claude, ChatGPT, Cursor or another MCP client. Chat apps connect with OAuth, and scripts use API keys. See [Agent API and MCP](/features/agent-api-and-mcp) for the setup.

## Related

- [Agent API and MCP](https://getkinlead.com/features/agent-api-and-mcp.md)
- [Lead groups and scoring](https://getkinlead.com/features/lead-groups-and-scoring.md)
- [AI LinkedIn outreach](https://getkinlead.com/features/ai-linkedin-outreach.md)

## Common questions

**Do I need Sales Navigator to automate LinkedIn lead search?**

No. A tool can use LinkedIn's standard search. A Sales Navigator seat is not required for that approach.

**How do I keep an AI from sending bad messages?**

Read the first messages from each campaign. Fix the reason line and the tone. Then let the campaign run.

**What should I do when a lead replies with a question I cannot answer?**

Answer the question directly. If you need time, say you will check and give a time. Do not send a pitch in place of an answer.

**Can I use the same message for every lead in a group?**

You can use the same structure, but the first line should name the reason that person is in the group. If the reason is the same for every lead, the group is too broad.

**How many leads should I review at once?**

Review a small batch you can check in one sitting. Fix the scoring rule before you add more leads.

[Create a Kinlead account](https://getkinlead.com/signup) or [book a call](https://getkinlead.com/demo) to talk through pricing and setup.
