MCP for HR and Recruiting: ATS and HRIS Connectors, Bias Laws and Human Decisions

MCP lets an AI assistant search candidates in your applicant tracking system, look up reporting lines and time off in your HR system, and draft offer letters and interview plans. As of November 2026, Greenhouse, Rippling, Gusto and Deel all document MCP servers. AI hiring laws in places like New York City and Illinois apply to how you use the output.

MCP (Model Context Protocol) is an open standard for connecting AI apps to other software through MCP servers. New to it? Read What Is MCP?.

This guide covers what MCP can do for HR and talent teams, which servers are documented, the legal limits and a starter stack. Facts were checked on 2 November 2026. This is not legal advice: employment law varies by place, so talk to your employment counsel or compliance lead before using AI in any hiring or employment decision.

What MCP can do in HR and recruiting

Useful first jobs are look-ups, reports and drafts:

• Pipeline reports. "How many candidates are in onsite stage for each open engineering role?"
• Interview prep. Pull a candidate's application and the job's scorecard to draft interview questions for the hiring manager.
• People look-ups. Find someone's manager, team or time-off balance.
• Onboarding drafts. Start a draft new hire record for HR to finish and approve.
• Policy Q&A. Answer employee questions from the handbook through a read-only server.

These save time on admin work. They should not rank, score or reject people without a human making the call.

Which HR and recruiting MCP servers exist

Greenhouse

Greenhouse's MCP overview says "Greenhouse MCP is currently in open beta" and is "Available to Core, Plus, and Pro subscription tiers." A Site Admin sets scopes in Dev Center, which cap what users can approve. It lists Claude, ChatGPT, Google Antigravity, Glean, Copilot Studio, Amazon Q and Grok as supported tools.

The Greenhouse MCP security FAQ is one of the clearest vendor security pages we have seen:

• It uses OAuth 2.0 with PKCE, so users sign in with Greenhouse and credentials never go to the AI tool.
• "Every DELETE endpoint is permanently blocked for MCP tools."
• Five actions "require explicit human confirmation before they run": rejecting an applicant, hiring an applicant, merging candidates, anonymizing a candidate and bulk-revoking user permissions.
• The Greenhouse Audit Log "captures detailed records of every API call made through an AI tool," tagged mcp_access, with the user, OAuth client, endpoint and time.
• "Greenhouse does not receive, process, or store the content of AI conversations." Your AI vendor's policies govern that.

Rippling

Rippling's MCP overview calls it "a first-party Model Context Protocol (MCP) connection." Actions run under the connected employee's own identity. Admins decide which tools each person can use through MCP Gateway policies and can revoke tools company-wide. Employees have no access until an admin creates a policy. See our Rippling MCP server profile.

Gusto and Deel

• Gusto. Claude's Gusto connector listing says it lets you "query and analyze your payroll and people data," and that "Eligible customers can also run payroll in Claude." See the Gusto MCP server profile.
• Deel. Deel's MCP docs show tools such as listOfContracts, getListOfPeople and createTimeOffRequest, with OAuth2 recommended. See the Deel MCP server profile.

What we could not confirm

We did not find enterprise HRIS or other major ATS vendor MCP documentation we could open and verify on 2 November 2026, apart from those above. Third-party integration platforms offer wrappers around many HR systems. Those add another vendor holding employee data, so review them first with our server risk checklist.

Legal limits on AI in hiring

New York City Local Law 144

NYC's Department of Consumer and Worker Protection explains Local Law 144. Employers and employment agencies may not use an automated employment decision tool unless it "has been subject to a bias audit within one year of the use of the tool." They must "Post a summary of the results of the bias audit" and give notice "10 business days prior to use." Enforcement began on 5 July 2023.

Whether an AI workflow built on MCP counts as an automated employment decision tool depends on how its output is used. Ask your counsel before an AI ranks or screens candidates for NYC roles.

Illinois Public Act 103-0804

Illinois amended its Human Rights Act in Public Act 103-0804, which "takes effect January 1, 2026." It makes it a civil rights violation for an employer to use artificial intelligence "that has the effect of subjecting employees to discrimination on the basis of protected classes," or "to use zip codes as a proxy for protected classes." It also requires employers to give notice when they use AI for recruitment, hiring, promotion, discipline and other covered decisions.

Everywhere else

Federal and state anti-discrimination laws apply to employment decisions however they are made. Other states and cities have passed or proposed AI-specific rules. Your employment counsel should keep a list for each place you hire.

Data protection inside HR

HR systems hold some of the most sensitive data a company has: pay, health-related leave, performance notes and background checks. Practical steps:

• Per-user sign-in. Use servers that act as the signed-in user, so access follows existing HR permissions.
• Narrow scopes. Start with read scopes, as Greenhouse lets Site Admins do in Dev Center.
• No cross-pollination. An AI with both HR and general chat access can move data between them. Read Cross-Server Data Leakage in MCP.
• Logs. Keep the vendor audit log and your AI client's logs. Our MCP Guide's audit logging article lists what a useful log contains.

Example prompts

• "How many candidates are in each stage for our open product roles, and which stages have candidates waiting more than ten days?"
• "Draft interview questions for this role based on its scorecard, for the hiring manager to edit."
• "Who reports to Maria, and who is out of office next week?"
• "Start a draft new hire record for the accepted offer, and list what HR still needs to complete."

None of these asks the AI to rank or reject a person. Keep it that way unless counsel has reviewed the workflow.

A starter stack

Plan approvals with our tool permission planner.

What to keep with people

• Screening, rejecting and advancing candidates.
• Offers, compensation and hiring decisions.
• Performance ratings, discipline and terminations.
• Accommodation and leave decisions.
• Payroll changes.
• Investigations and complaints.

The MCP specification says applications "SHOULD" present confirmation prompts so a human stays in the loop, per the 2026-07-28 tools specification. Greenhouse builds that into its highest-risk actions, which is a good model to copy.

Next step

Our MCP readiness audit reviews your HR systems, permissions and legal requirements before you connect an AI. If you need a custom read-only server, for example over your handbook or job architecture, see MCP development. For a comparison list, read Best MCP Servers for HR.

Frequently asked questions