In episode one, we covered whether you should use AI as a recruiter. This article covers the practical side in episode two: five things you can do with the Pinpoint MCP today, with the exact prompts to try.

The good news: there's nothing new to learn. No training, no new interface. You use the AI tools you already have (like Claude or Microsoft Copilot) and just describe what you need.

What makes this possible is the Pinpoint MCP. MCP stands for Model Context Protocol, a secure bridge that connects your AI assistant directly to your live Pinpoint account. Connect once, sign in with your normal Pinpoint login, and it can read your pipeline, rank candidates, draft outreach, and take actions. It acts as you, with your permissions. Anything actions that changes data, you approve and everything is logged.

To help get you started, we've put together a library of 25 prompts you can use right now. The Pinpoint MCP Prompt Template Library covers every use case below and more, organized by task type.

5 ways to use an MCP in TA today

1. Check your full pipeline in seconds

It's the simplest thing in this article, and often the most useful. Instead of opening Pinpoint and scanning through each stage, you just ask. It doesn't just give you a count either. It flags who applied in the last day, so you know about overnight candidates before you've even opened a tab.

Try it yourself

1. Open Claude Desktop (or Copilot) with your Pinpoint connection active
2. Paste this prompt, swapping the bold parts for your real role:

"Give me a snapshot of my [Account Executive] pipeline: how many candidates at each stage, who applied in the last day, and who's at interview."

One sentence, and you know exactly where your day starts.

2. Shortlist candidates against the brief

If you have multiple applicants for one role, instead of opening each resume and comparing manually, you can ask the MCP to rank them against the job brief. It reads every resume alongside the job description and returns a ranked table with a one-line verdict and the single biggest strength or gap for each candidate. What used to take an hour takes seconds, and it's fairer too, because every candidate gets measured against exactly the same brief.

Try it yourself

1. Make sure the job brief is attached to the role in Pinpoint
2. Use this prompt:

"Read the job brief for [role name] and rank my applicants against it. For each, give a one-line fit verdict and the single biggest strength or gap. Who should I prioritize?"

Adjust the wording if you want to focus on specific criteria from your intake note.

3. Compare your finalists

Picture this: you have two candidates at final stage and a debrief tomorrow. The prep normally involves re-reading resumes, scrolling back through interview notes, and tracking down the hiring manager's intake brief.

Or you can just ask.

I asked it to pull resume highlights, the panel's interview feedback, and the intake note into one side-by-side comparison and tell me the key decision. It returned a table and framed the actual trade-off in terms of what the hiring manager had told us they cared about.

Try it yourself

"Compare [Candidate A] and [Candidate B] for [role]. Pull resume highlights, the interview feedback the team left, and how they map to the intake note. Put it in a table and tell me the key decision."

The part that surprises people: it pulls in the intake note. That's what a spreadsheet comparison can't do.

4. Draft, task, and notify in one go

Up to this point, you've been asking for information. This is where you start taking action. I asked it to draft an interview invitation, create a follow-up task, and leave the hiring manager a note, all at once, using the candidate's real data from Pinpoint. It showed me everything before touching anything. I approved, and all three things happened.

Try it yourself

"[Candidate name] is my front-runner. Draft a short, warm email inviting them to a final interview, add a task for myself to chase the panel's feedback by [day], and leave the hiring manager a note that they're progressing. Show me before anything sends."

5. Set your morning triage to run itself

Everything above required you to be there. This one doesn't.

Claude Desktop has a scheduling feature (on paid plans) that lets you set a recurring prompt. I set a weekday 7:00 a.m. digest: new applicants across all open roles since yesterday, grouped by job, with a one-line verdict on each. It runs while I sleep. By the time I open my laptop, the overnight triage is done.

Try it yourself

"Every weekday at 7am, summarise the new applicants across the roles I'm hiring on since yesterday, grouped by job, with a one-line 'worth a look?' each."

Note: scheduled tasks are a Claude Cowork feature, available on paid Claude plans. The Pinpoint connection works the same regardless.

Key takeaways

• You don't need to learn a new tool or change the way you work. You simply ask what you need in plain English. That's the whole point.
• A lot of recruiting work is spent pulling information together before you can act. Reading resumes, gathering interview notes, tracking down the hiring manager brief. The MCP does that heavy lifting so you can spend more time making decisions and less time searching for context.
• It connects actions together. Instead of asking for information, then updating a candidate, then sending a note to a hiring manager separately, you can do all of that from a single prompt. You're still in control, but a lot of the admin disappears.

What to do this week

1. Connect your hiring data to your AI tool

1. Ask your Pinpoint admin to enable MCP access in Settings → Company → Preferences → AI Features.
2. In Claude Desktop or Microsoft Copilot, add your Pinpoint MCP URL: https://<your_subdomain>.pinpointhq.com/mcp
3. Sign in with your normal Pinpoint login and click Allow.

You can read more about the set up in our help article.

2. Try your first MCP workflow

Open the Pinpoint MCP Prompt Template Library, pick a prompt that fits a workflow you find repetitive, and try it. You'll quickly get a feel for where it can save you time.

‍