Build an AI Recruiting Agent with LinkedIn Jobs and Email

Recruiting is a numbers game wrapped in a research problem. You look at a lot of profiles to find a few good fits, and then you write personalized messages to each one. Both parts take time. An AI agent can handle most of both.

What Recruiting Actually Involves

There are four main activities in a typical recruiting workflow: sourcing, researching, outreach, and follow-up.

Sourcing means finding candidates in the first place. LinkedIn is the default. So are job boards, referrals, GitHub, and conference speaker lists. A recruiter might scan 50 profiles to find 10 worth contacting.

Researching means reading those profiles and forming an opinion. What did this person build? Where did they work before? What do they post about? The goal is to understand whether the role fits them, and to find one specific thing to reference in your message.

Outreach is where most recruiting falls flat. Generic InMail reads like a form letter. A message that references a specific project, a paper they published, or something they wrote gets read. Writing that message takes 10 minutes per candidate if you’re doing it right. Multiply by 10 candidates and it’s an afternoon.

Follow-up is easy to forget. Candidates go quiet. A second message a week later often converts. Tracking who you emailed, when, and whether they responded is its own small project.

Where the Time Goes

The bottleneck isn’t judgment. You still need a human to decide whether a candidate clears the bar. The bottleneck is the mechanical work that happens before and after that decision: finding profiles, reading them, writing messages, sending them, and watching for replies.

That’s mostly research and composition. Two things AI agents are good at.

What an Agent Can Do

An agent running a recruiting workflow looks something like this:

  1. Search LinkedIn for candidates matching a job description.
  2. For each result, pull the profile and look up any public work (GitHub, personal site, published writing).
  3. Compose a message that references something specific from that research.
  4. Send it from a dedicated outreach address.
  5. Check the inbox for replies and surface anyone who responded.

The agent doesn’t decide who to hire. That’s still yours. But it can hand you a shortlist of warm replies instead of a cold spreadsheet of profiles.

The “personalized at scale” problem gets easier too. An agent can research 20 candidates in the time a human researches 2, and each message can still reference something real. That’s not magic. It’s the same research process, done faster.

Setup

Connect AgentPatch to your AI agent to get access to the tools:

Claude Code

claude mcp add -s user --transport http agentpatch https://agentpatch.ai/mcp \
  --header "Authorization: Bearer YOUR_API_KEY"

OpenClaw

Add AgentPatch to ~/.openclaw/openclaw.json:

{
  "mcp": {
    "servers": {
      "agentpatch": {
        "transport": "streamable-http",
        "url": "https://agentpatch.ai/mcp"
      }
    }
  }
}

Get your API key at agentpatch.ai.

The Tool Stack

AgentPatch gives the agent everything it needs through a single connection:

linkedin-jobs-search (75 credits): searches LinkedIn for candidates by role, location, and keywords. This is your sourcing step. Tell the agent what you’re looking for and it returns a list of profiles to work from.

scrape-web (200 credits): pulls and renders any URL. Use it to read a candidate’s GitHub, personal site, portfolio, or published work. This is where the personalization material comes from.

send-email (100 credits): sends from an address you’ve claimed through AgentPatch. Before sending for the first time, run claim-email-address (5,000 credits, one-time) to get a dedicated inbox. That address handles both outbound messages and inbound replies.

check-inbox (25 credits): reads replies to that address. The agent can scan for responses, find candidates who replied, and flag them for your review.

No LinkedIn API credentials. No Resend account. No SMTP config. One API key handles the whole stack.

A prompt to get started might look like:

“Search LinkedIn for senior backend engineers in New York with Go experience. For the top 10 results, look up their GitHub profiles and any public projects. Write a personalized recruiting email for each one referencing something specific from their work. Send the emails and let me know when replies come in.”

The agent runs that sequence start to finish. You review the replies.

Wrapping Up

Recruiting involves a lot of work that doesn’t require a recruiter: searching profiles, reading them, writing messages, tracking responses. An AI recruiting agent handles that layer while you focus on the part that matters: deciding who to talk to and what to offer them. Visit agentpatch.ai to connect the tools and run the workflow.

← Back to blog