How to source candidates with Claude

Sourcing has always been a tab-juggling job: LinkedIn in one window, a contact database in another, an email finder in a third, a spreadsheet holding it all together. AI assistants change the shape of that work. With MCP (Model Context Protocol), Claude and ChatGPT can call sourcing databases directly — which means you describe the role once and the assistant does the searching, enriching, and verifying for you.

This guide walks through the full workflow with Scout Mesh, an MCP gateway that connects your assistant to nine sourcing data sources through one endpoint.

What you need

• Claude (Pro, Team, or Enterprise) or ChatGPT with connector support
• A Scout Mesh account — new accounts start with $5 in free credits, which is enough to run a real search end to end

You do not need subscriptions to Lusha, People Data Labs, or any other database. The gateway carries the access; you pay each provider's unit price only for the calls you make.

Step 1 — Connect the gateway

Add the Scout Mesh MCP server to your assistant as a custom connector and sign in. From that point, every source in the mesh — People Data Labs, Lusha, ContactOut, GitHub, Hugging Face, prog.ai, TheirStack, Bouncer, and Exa — appears as tools the assistant can call. This is the only integration step.

Step 2 — Describe the role, not the query

Instead of building Boolean strings, tell the assistant what you actually need: "Find me five senior Go engineers in Amsterdam who have worked at product companies, and get verified emails for the best three." The assistant picks the right source for each part of the job — code-activity search for engineers, contact databases for emails, the open web for background research.

Step 3 — Search wide, spend narrow

A good assistant-driven search starts with the free and cheap layers: GitHub and Hugging Face searches cost nothing, and Exa web research costs $0.007 per query. The assistant builds a longlist there, then spends on enrichment — a full profile or verified email — only for candidates who survive your review. You see each cost before it is incurred.

Step 4 — Verify before outreach

Every email that makes the shortlist should pass through independent verification before it reaches your sequences. Bouncer checks deliverability at $0.008 per address — batch-verifying a 50-person list costs $0.40 and protects your sending domain from bounce damage. We wrote up why this matters in the email verification guide.

Step 5 — Hand off to outreach

The output is a clean shortlist — names, roles, evidence, verified contacts — assembled in one conversation. Export it to your ATS or outreach tool of choice; the sourcing part is done.

Why one gateway beats separate tools

Each database is strong somewhere: People Data Labs for professional depth, prog.ai for real code skills, Lusha for phone numbers, ContactOut for LinkedIn-to-email. Wiring them into your assistant one by one means separate subscriptions, separate auth, and no way to compare sources. A gateway makes them one bill and one conversation — and lets the assistant cross-check one source against another, which is where the quality gains actually come from.