From monthly reports to a two-minute dashboard: How a TA Ops lead transformed her week with Greenhouse MCP
Jump Links placeholder, DO NOT REMOVE
Key highlights:
- Greenhouse MCP is a governed way to connect your own AI tools directly to Greenhouse Recruiting, working inside the permissions and audit trail you already have.
- One TA Ops lead built six or seven repeatable workflows: bulk approval-chain updates, data integrity reviews, a company-wide job post catalog and a live KPI dashboard.
- The KPI dashboard pulls ATS data and refreshes in under two minutes, moving reporting from monthly to every two weeks.
- The biggest wins came from the manual, on-repeat work she already did every week, not from anything transformational on day one.
A Greenhouse TA Ops lead on how she rebuilt her week with MCP, in her own words.
So when Greenhouse launched Greenhouse MCP, a governed way to connect your own AI tools directly to Greenhouse Recruiting, I paid attention. It isn’t a feature I waited for someone to ship. It’s a connection layer I pointed my own AI at, and it’s been the most impactful change to my day-to-day in those three years.
I’ll be honest about my first reaction, though: I read the list of permissions and had no idea how any of it translated into the work I actually do. So I did what I’d tell any TA Ops person to do: I started asking Claude. “Using the Greenhouse MCP, can I do this? Can I do that?” That was the unlock. From there I built six or seven workflows that changed how I spend my week.
Here’s what I built, and why each one exists.
One thing worth saying up front, because it’s what made me comfortable building all of this: MCP works inside the permissions I already have. Every action runs through the same access controls and audit trail as anything else I do in Greenhouse Recruiting. I’m not reaching around the system. I’m working through it.
Bulk permissions and approval chain updates
This is the one that saves me the most time.
We’ve had a lot of organizational movement lately: business partners shifting from one team to another, approval chains that need to follow them. Before MCP, that meant opening every affected job one by one, updating the approver, saving, moving on. It’s the kind of work that eats an afternoon and feels invisible to everyone except you.
Now I open Claude Cowork and say something like, “Replace Mary with Danielle as the approver for all marketing roles across Greenhouse Recruiting.” It does it, working through the same permissions I already have, so the change is governed the same way it would be if I’d clicked through every job by hand. A few hours of work, one prompt. I’ve run this every time we’ve had a personnel change since, and it hasn’t gotten old.
Job post catalog for the whole company
This one started as a one-off and grew into a real system.
Every time we make a hire, we now extract the job post that hire was associated with, name it after the person who got the offer and save it into a shared Google Drive. That gives our legal team, People Operations and leave and benefits folks instant access to the job description for any employee, whenever they need it. No more “can you send me the JD for the role X was hired into?” emails three times a quarter.
We actually started this with a backlog of about 1,300 job posts to save. MCP handled the second half of that project once we got access, which saved us close to 20 hours of manual work, and now it runs monthly on its own.
It’s a small thing on the surface. It’s a real time saver across multiple teams in practice.
Reviewing for data integrity
Data integrity is one of those areas where the work is critical because the accuracy of our reporting hinges on it, but it requires a level of attention to detail that requires us to drop everything and focus.
When opening a new job I review job custom fields for accuracy and completeness – a process that takes 15 minutes of deep concentration per job. During high volume hiring periods this becomes tedious and disruptive to other strategic work.
MCP turned it into a call that can be run to compare inputs against similar jobs and flag anything that is incomplete or unexpected.
The confidence we have in our data integrity remains the same, but how much time we dedicate to maintaining it is dramatically reduced.
From ATS data to a live KPI dashboard in two minutes
This is the workflow I’m proudest of.
We built a live artifact in Claude that pulls data through Greenhouse MCP and turns it into a full recruiting KPI dashboard, complete with the context behind the numbers. It covers core metrics like time-to-fill, hires per recruiter, offer acceptance rate and candidate sentiment, plus the supporting signals we actually need to manage the function: hires by recruiter, department, level, source and office; outstanding offers; average days live for active roles; roles kicking off; and fraud risk flags for candidates who have reached our Assessment milestone.
The whole thing refreshes in under two minutes.
Before this, we were reporting on basic KPIs once a month because pulling the data together took too much manual work. Now we review KPIs and supporting contexts every two weeks.
What used to take about two hours now takes minutes, and the team gets a much more complete picture of what is happening.
That gives us not only a faster report, but also the ability to react and pivot in real time.
Quality of assessment (still exploring)
This one is unfinished, and I’m sharing it anyway because I think it has legs.
We took every hire from 2025 and used MCP to build a few indicators of how well we were assessing candidates, not how the hires turned out. Average scorecard score, treating “definitely not” as 1 and “strong yes” as 5. Average scorecard attribute confidence, so we could see how consistently we were vetting candidates against the things that actually matter for the role. And then qualitative reads, having Claude summarize key themes in scorecard feedback or flag patterns worth a second look.
I want to be clear about what this is and isn’t. We’re not measuring quality of hire. That would mean tying these candidates to post-hire outcomes like performance or tenure, and we haven’t gone there, both because quality of hire is genuinely hard to pin down and because it only means something once the business has defined what “quality” is for a given role, which most haven’t.
What we can measure right now is the quality of our own assessment: whether our interviewers are aligned, and whether a decision rested on real signal or moved to an offer on thin evidence, which usually means we hadn’t defined the attributes that mattered.
We didn’t see enough variance to roll it out organization-wide, but the workflow exists. If you can read scorecard feedback across every hire, you can start asking real questions about how your interview process is performing, not just whether people are filling things out.
Smaller pulls that add up
Plenty of smaller asks come through that used to be tedious. A colleague needed a list of active interviewers across every role for a Notetaker rollout. Before MCP, that meant going into each job and interview kit to record names by hand, at least half a day of work. With MCP, it’s one prompt. Those individual time savings stack up.
What I’ve learned about using AI in TA operations
The MCP rewards specificity
When I started, I asked broad questions and got vague answers back. The fix was learning to reference specific job IDs, candidate IDs, application IDs wherever possible. If I need to do something across a known set of records, I’ll go into the Greenhouse Recruiting report builder first, pull the relevant data, export it and bring it into the MCP. That way Claude isn’t guessing what to look up. It’s referencing.
Ask Claude what it can do
This sounds obvious, but it’s the single most important thing I did. The permissions list doesn’t tell you what’s possible in practice. The conversation does.
Start with the work you do every week
The biggest wins for me were the workflows I was already doing manually on repeat. MCP doesn’t have to be transformational on day one. It can just take an hour-long task and make it a five-minute task. Do that six or seven times and your week looks different.
Greenhouse MCP is now in open beta. Check out how other leading hiring teams are using Greenhouse MCP for success. Then, explore our other AI tools that keep hiring structured, explainable and human-led.
FAQs
What is Greenhouse MCP?
Greenhouse MCP (Model Context Protocol) is a governed connection layer that lets you connect your own AI tools directly to Greenhouse Recruiting. It works inside your existing permissions, so anything you do through it runs through the same access controls and audit trail as anything else you do in Greenhouse.
What can TA operations teams actually do with it?
Common uses include bulk permission and approval-chain updates, GDPR-compliant candidate data exports, building a searchable catalog of job posts and pulling ATS data into a live KPI dashboard. The pattern that works best is starting with the manual, repetitive tasks you already do every week.
Does using AI through MCP change who can access candidate or hiring data?
No. MCP respects the permissions you already have. It doesn’t grant new access or reach around your controls; it acts within them, which is what makes actions like bulk updates safe to run this way.
How do I get started with Greenhouse MCP?
Greenhouse MCP is now in open beta. The fastest way to learn what’s possible is to connect it and ask your AI tool directly what it can do with it.
.png)
