MCP for HR: What It Is and Why It Matters

Every HR vendor claims to be “AI-first” now. Far fewer can actually connect to the AI agent your team is already using. Model Context Protocol — MCP — is the standard that closes that gap, and it’s quietly becoming the most important line item on HR software evaluation checklists.

This guide explains what MCP is, why it matters for HR specifically, which platforms support it today, and what to do with that information.

What MCP actually is

MCP is an open protocol released by Anthropic in November 2024 that lets AI assistants connect to external software. It standardizes the way tools expose data and actions to AI models, so any MCP-compatible AI client can talk to any MCP-compatible server.

Before MCP, every AI integration was bespoke. ChatGPT had plugins. Claude had connectors. Each tool built one-off integrations with each AI. Developers rewrote the same glue code for every combination.

With MCP, a vendor writes one server. Every AI client — Claude, Cursor, ChatGPT Agent, Codex — can use it. That’s why adoption is accelerating fast across enterprise software, and why the biggest story in HR tech for 2026 is which platforms ship MCPs and which don’t.

Why HR needs MCP more than most

HR data is scattered by design. HRIS holds employee records. Performance platforms hold reviews. ATS holds candidate data. Payroll holds comp. Culture tools hold survey responses. Each system is a silo.

Cross-system questions are the norm, not the exception: “Who’s starting next week and needs onboarding scheduled?” “Which managers are running behind on reviews?” “What did our last engagement survey say about the engineering org?”

Without MCP, answering any of those means either exporting CSVs and pasting them into ChatGPT, or getting IT to build a custom pipeline for every question. With MCP, an AI assistant can query the right systems live, in natural language, using only the permissions the employee running the query already has.

The Wikipedia page on MCP describes the result accurately: AI assistants stop being isolated chatbots and start being operational tools that work across the systems you already use.

What MCP actually lets HR teams do

Concrete examples, not hypotheticals:

Onboarding coordination. “Who’s starting next week? Draft a Slack message to each of their managers with onboarding links, and schedule their first 1:1 for day two.”
Performance cycle status. “Which managers haven’t started their reviews yet? Send them a reminder.”
Survey analysis. “Summarize the latest pulse survey responses by department and flag teams with declining engagement.”
Feedback loops. “Based on who worked with Alex on the Q1 launch, draft a feedback request for each collaborator.”
Calibration prep. “Show me managers whose rating distributions look unusual compared to the company average.”

Each of these requires pulling structured data from HR systems and reasoning over it. None of them work well if the AI can only read a PDF export.

The current state of HR MCP support

The landscape as of early 2026 falls into three buckets.

Platforms with official MCPs:

HiBob (beta) — people data, time off, tasks
Windmill — org chart, performance reviews, 1:1s, pulse surveys, feedback, recaps
Gusto — payroll and benefits data
Checkr, Manatal, Draup, Leonar — narrower scopes within their categories

Platforms with partner-built or unofficial MCPs:

Workday — accessible via Workato and other middleware
BambooHR — only community-built servers exist; no official offering yet

Platforms without MCP support:

Lattice, Culture Amp, 15Five — no public MCP servers
Most ATS platforms beyond SmartRecruiters and a few others
Most benefits administration tools

The absences matter. A tool without an MCP (or a published roadmap commitment) will become harder to use as AI assistants become the default interface for white-collar work. Some HR leaders are already deprioritizing vendors that lack public APIs.

How MCP security actually works

The scariest-sounding part of MCP is the least novel. MCP is a protocol, not a security model. Good implementations use OAuth and inherit your existing permissions: if you can’t see something in the dashboard, the MCP can’t return it either.

The risks to actually watch for:

Unapproved actions. An agent reading data is low-risk. An agent sending a Slack message or creating a Notion doc on your behalf is higher risk. Set write and delete tools to require explicit approval, especially in unattended scheduled tasks.

Untrusted third-party MCPs. The biggest exploits in the MCP ecosystem have come from installing community servers without reading the code. An MCP server is just a program you’re running — treat it with the same caution you’d treat any other software.

Vendor data handling. Check for zero-data retention and confirm whether your data can be used for training. Most enterprise plans now default to no-training, no-logs, but confirm in writing.

What to do with this information

Three practical takeaways for HR leaders.

Audit your existing stack. For each tool in your HR stack, ask: does it have an MCP? A public API? A published roadmap for AI integration? Tools without any of those are becoming dead weight. Windmill’s integrations page is one reference for what MCP access looks like in practice.

Bake MCP into procurement. When evaluating new HR software, require vendors to answer specific questions: Do you expose an MCP server? Which tools does it expose? What’s your OAuth model? What’s the authorization scope? “We have an API” is not the same answer as “we have an MCP.”

Use the MCPs you already have. If your current HRIS has an MCP (or your performance platform does — Windmill’s MCP is the only one in that category as of writing), set it up. The workflows it unlocks — automated onboarding handoffs, scheduled survey recaps, live org chart queries — require almost no custom engineering.

MCP is still early. The vendors who ship one now will define the default HR tech stack for the next decade. The vendors who don’t will be working from behind.

Frequently Asked Questions

What is Model Context Protocol (MCP)?

MCP is an open standard, created by Anthropic in late 2024, that lets AI assistants connect to external software. Instead of every AI tool building custom integrations with every app, MCP defines one protocol that any AI client can use to talk to any MCP-compatible server. For HR, it means your AI assistant can read and act on data in your HR platforms without a bespoke integration.

Which HR platforms support MCP?

As of 2026, HiBob offers an official MCP server in beta, and Workday has partner-built MCP servers through providers like Workato. BambooHR has no official MCP but has several community-built servers. Windmill, Gusto, Checkr, and Manatal also offer official MCPs. Lattice, Culture Amp, and 15Five have not released MCP servers.

Why does MCP matter for HR teams?

HR data lives across many tools — HRIS, performance platforms, ATS, payroll, surveys. Without MCP, AI assistants can't reason across that data without custom integration work. With MCP, an AI can answer questions like 'Who's starting next week and needs onboarding?' by pulling from multiple HR systems in a single conversation, using only the permissions the employee already has.

Is MCP secure enough for HR data?

MCP itself is a protocol, not a security model. Good MCP servers use OAuth authentication and inherit the user's existing permissions — if you can't see a document in the dashboard, the MCP won't return it either. The risks to watch for are agents taking unapproved actions, and downloading untrusted third-party MCP servers that could exfiltrate data. Enterprise plans with zero-data retention address most vendor-side concerns.

Should I pick HR tools based on MCP support?

Increasingly yes — not because MCP itself is magic, but because MCP support signals a vendor is committed to open, AI-ready architecture. Vendors without MCPs or public APIs will become harder to use as AI assistants become the default interface for knowledge work. Ask for a roadmap commitment even if MCP isn't shipping today.