Recruiting MCP Servers 2026: Resume Parsing, Matching, ATS

To parse resumes and score candidates from Claude, ChatGPT or Cursor, use a parsing MCP server such as HireLayer, RChilli or Affinda. To act on your hiring pipeline, use your ATS's own server: Greenhouse, Ashby, Workable and Teamtailor all host one. Job seekers have their own set, from Indeed to AceCV. This guide lists the recruiting MCP servers that exist today, what each one can do, how it signs you in and what it costs, so you can pick the right one in a few minutes.

A word on bias: HireLayer is our product and it is on the list. Every other fact comes from the publisher's own page, docs or GitHub repository, checked on 7 October 2026. The links are in the sources section at the end.

The short list

Server	What it does	Sign-in	Cost
HireLayer	Parse resumes, extract job criteria, match and rank candidates, normalize skills	OAuth (hosted) or API key (local)	50 free calls a month, then from €24 for 500
RChilli MCP Hub	27 read-only tools: parsing, taxonomy lookups, resume-to-job scoring	OAuth	100 trial credits, Enterprise on quote
Affinda	About 75 tools to set up and run document extraction workspaces	OAuth 2.1	Trial with 200 credits, prices on request
Pin	AI sourcing: jobs, candidate lists, accept or reject	OAuth 2.1	Included in paid Pin plans, from $99 per user a month
Greenhouse	A subset of the Harvest API on your own Greenhouse data	OAuth 2.0	Open beta on Core, Plus and Pro
Ashby	17 tools: candidates, interviews, feedback, pipeline stages	OAuth per user	Open beta on all plans
Workable	Over 90 recruiting and HR tools	OAuth 2.0	All plans
Teamtailor	Read, create and delete actions, set per role	OAuth 2.1	Paid add-on

What a recruiting MCP server is

The Model Context Protocol (MCP) is the standard way to give an AI assistant new tools. An MCP server publishes a list of tools, such as "parse this resume" or "move this candidate to the next stage", and the assistant decides when to call them while it answers you. Claude, ChatGPT, Cursor, VS Code and most coding agents speak MCP, so one server works in all of them.

Recruiting servers fall into three groups, and they rarely overlap. Parsing and matching servers read documents you give them and return structured data or scores. ATS servers read and change the records in your applicant tracking system. Job-seeker servers work for the candidate: they search jobs, edit a CV and sometimes apply. Most teams that screen candidates with an assistant end up using one server from the first group and one from the second.

Parsing and matching servers

HireLayer

HireLayer's server exposes one tool per API: parse_resume turns a resume file into JSON, extract_job_criteria turns a job description into weighted criteria, match_candidate scores one candidate against one job criterion by criterion, rank_candidates orders up to 10 candidates for a job with a rationale each, and resolve_skills maps free-text skills to a taxonomy of 11,061 skills. It also ships prompts for common flows, like screening a batch of CVs against a job.

The hosted server lives at https://hirelayer.co/mcp and signs you in with OAuth, so there is no key to paste. To read files from your own disk, run it localy with npx -y hirelayer-mcp and an API key. The code is open source under the MIT licence. Every successful call uses one credit; the free plan includes 50 a month without a card and paid plans start at €24 for 500. Resumes are processed in Paris, France. Setup steps for each client are on the HireLayer MCP page.

Best for: teams that want parsing, matching and ranking from one server with published prices.

RChilli MCP Hub

RChilli hosts the largest parsing toolset: 27 tools, all read-only according to its page. They cover resume and job parsing ( extract_resume_data, extract_job_data), skill and job lookups in O*NET and ESCO, redaction and reformatting, scoring a resume against a job (score_resume_against_jd), searching an index of documents, and agent-style tools such as detect_bias and generate_interview_questions. Sign-in goes through an RChilli account with OAuth. A trial gives 100 free credits; the Enterprise plan is on quote.

Best for: teams already on RChilli, or who want taxonomy lookups next to parsing.

Affinda

Affinda's connector is closer to an admin console than a parser: around 75 tools manage organisations, workspaces, document types, fields and API tokens. One call, create_recruit_workspace, sets up a recruiting workspace with resume, job description, redacted resume and resume summary document types. There is one endpoint per region (Australia, US and EU) and sign-in uses OAuth 2.1. Extraction is billed per page; the trial includes 200 credits and prices are on request.

Best for: Affinda customers who want to configure and run extraction from an assistant.

Pin is a sourcing product, not a parser. Its 10 tools create a job, fetch the candidates Pin found, accept or reject them and adjust the search. The server is included at no extra cost in paid Pin plans, which start at $99 per user a month billed yearly.

Best for: recruiters who source candidates with Pin and want to drive it from Claude or ChatGPT.

Open-source parsers

Small open-source servers exist, such as resume-parser-ai-mcp (MIT, installed with pip). Read the code before relying on one: this one takes resume text rather than files and finds fields with regular expressions and keyword lists, wich works for clean text but not for real CV layouts. They are fine for experiments and demos.

ATS servers: act on your hiring pipeline

If your question is "which candidates are waiting for feedback?" or "move Jane to the onsite stage", you need your ATS's own server. All four below are hosted by the vendor and sign each user in with OAuth, so the assistant only sees what that user can see in the ATS.

Greenhouse: a subset of the Harvest API as tools, in open beta on Core, Plus and Pro. Delete calls are blocked and five high-stakes actions, including reject and hire, ask for confirmation. Hosted in the US only.
Ashby: 17 documented tools, from get_candidate to change_application_stage, in open beta on all plans since June 2026. An admin turns it on first.
Workable: over 90 tools across jobs, candidates, offers and HR data, on all plans.
Teamtailor: a paid add-on where admins choose, per role, whether the assistant can read, create and modify, or delete. Every action is logged.

Lever and Bullhorn do not publish an official server yet; third-party connectors exist for both. For Greenhouse there is also a community server, open-greenhouse-mcp, that runs localy with your own API credentials.

Servers for job seekers

These work on the other side of the table. Indeed's server, in beta and for Claude only, searches jobs and reads your Indeed resume. AceCV edits and reviews your CV (Pro plan, $14.90 a month). Resumly tailors resumes and can apply for you, billed per action, for example $0.05 to parse a resume and $0.50 per submitted application. Workopia searches jobs and tailors a resume for free. LoopCV has announced a server but it is still on a waitlist.

How to choose

Start from the job the assistant has to do. Reading CVs and job descriptions and turning them into scores is a parsing and matching job. Changing records is an ATS job. Many recruiting teams connect both: the assistant parses and ranks the new applicants with a parsing server, then writes a note or moves the shortlist in the ATS.

Then check four things before you connect a server:

What it can change. Read-only servers are safe to try. For servers that write, prefer ones that ask for confirmation or let an admin limit actions per role.
How it signs in. Hosted servers with OAuth keep keys out of config files. Local servers with an API key can read files on your disk, which is handy for a folder of CVs.
Where the data goes. Candidate data is personal data. Check the region of the server and of the service behind it, and whether files are stored.
What one call costs. Assistants call tools more often than you expect, so a published per-call price makes the bill predictable.

Finaly, keep a person in the loop. Scores and rankings help a recruiter decide faster; under the EU AI Act, systems that rank candidates are high-risk and need human oversight. Our EU AI Act checklist for recruitment covers what that means in practice.

Try the live demo Read the API documentation View pricing

Frequently asked questions

Is there an MCP server for resume parsing?

Yes. HireLayer, RChilli and Affinda all host one. HireLayer and RChilli expose parsing and scoring as tools you can call directly; Affinda's server mostly configures extraction workspaces. Small open-source servers also exist but are simpler.

Can ChatGPT or Claude rank candidates against a job description?

Yes, with a matching server. With HireLayer, the assistant calls extract_job_criteria on the job description and rank_candidates on up to 10 resumes, and gets a score and a rationale for each candidate. RChilli offers score_resume_against_jd for one resume at a time.

Which ATS have an official MCP server?

As of October 2026: Greenhouse, Ashby, Workable and Teamtailor host official servers. Lever and Bullhorn rely on third-party connectors. SmartRecruiters publishes an MCP server for its developer docs, not for live ATS data.

Are recruiting MCP servers free?

The server itself is usually free; you pay for the service behind it. HireLayer includes 50 free calls a month, RChilli and Affinda offer trials, and the ATS servers come with the ATS plan or as an add-on.

Is it safe to send candidate data to an MCP server?

Treat it like any other processor of personal data: check where it runs, whether files are stored, and what the server is allowed to change. OAuth sign-in and per-role permissions limit what the assistant can do on your behalf.

Sources and further reading

Louis Desclous

Published on October 7, 2026 · Reading time: 8 minutes