HireLayer: Resume Parsing and Candidate Matching APIs

HireLayer is a French recruiting API that turns a CV file into normalized JSON and, through four further endpoints, turns job descriptions into weighted criteria and scores and ranks candidates against them. The buyer is a software team that wants parsing and matching inside an applicant tracking system, job board or HR tool. Finance and operations documents are out of scope.

Company and ownership

HireLayer was called OnlineResumeParser until 2026, when it took its current name and moved to hirelayer.co. The legal notice names Louis Desclous, Le Plessis-Robinson near Paris, as publisher, and the privacy policy names him as data controller for accounts and billing. The site is published by an individual; no company registration number, headcount or funding is listed. It is not the job-search app at hirelayer.ai, which the vendor states is a separate company. The vendor reports 2M+ API requests processed since 2024 and a parse success rate of about 99.7%, defined as the share of valid CVs for which the API returns data. Field accuracy is a separate question, covered under Evidence.

Where HireLayer fits

HireLayer is a narrow tool. It handles one document type, the CV, and the recruiting workflow around it. Its own Wink case study shows the intended shape: an applicant tracking system calls CV Extract and Skills when a CV arrives, and Job Extract, Match and Rank when a job opens, with the ATS staying the system of record. Wink is an applicant tracking system for SMEs, agencies and public employers, based at Station F in Paris; it is the only customer named on the site.

Against general resume parsers such as Affinda, Textkernel or RChilli, HireLayer's product page and its Affinda comparison stress the following:

• Normalized vocabularies. Experience in five brackets, education on the European Qualifications Framework (Level 1 to 8), languages on CEFR levels A1 to C2, contract types (permanent, fixed-term, temporary, internship, apprenticeship, freelance), and French ROME occupation codes.
• Matching with explanations. Job Extract turns a job ad into criteria with weights and a mandatory flag; Match returns a score from 0 to 1 with a status and a reason per criterion; Rank orders up to 10 candidates with a score and one rationale each.
• Candidate photo. CV Extract detects the photo in a CV, crops it and returns it as an image URL that expires after 10 minutes when do_not_store_data is set. Affinda returns a headshot too, so this is parity rather than an edge.

One limitation is easy to miss: Job Extract criteria, Match explanations and Rank rationales are generated in French whatever the input language, and the mandatory flag is informational and does not change the score.

It is a poor fit for anyone who needs invoices, contracts or forms processed (see the wider IDP vendor directory), for teams that must keep CVs on their own infrastructure, and for front-end flows where a candidate waits on the parse: HireLayer's own benchmark puts a parse at 22 to 31 seconds.

Products

CV Extract takes a multipart upload of up to 6 MiB in 13 formats, including scanned and photographed CVs as JPG, PNG or BMP, and answers synchronously; a V2 callback contract remains for older integrations. Documents that are not a CV, unreadable or empty return a classified 422 error instead of an empty profile. Beyond the REST API, HireLayer ships an MCP server (OAuth or API key, open source under MIT, published on npm as hirelayer-mcp) for Claude, ChatGPT, Cursor and VS Code, and recipes for Make and Power Automate. HireLayer does not name the models behind the parser; the product page speaks of an "in-house AI pipeline" that runs in the EU, while the privacy policy states that CV Extract may run on AWS Lambda and Amazon Bedrock and that the four other APIs use Amazon Bedrock, without naming a Bedrock region.

Deployment, security and pricing

HireLayer is hosted only. Parsing runs in Paris, where uploaded files and extracted photos are also stored; there is no region choice, no self-hosting and no on-premises option, which HireLayer itself lists as a gap against Affinda. Retention can be switched off per request with do_not_store_data=true. On certifications the vendor states that ISO 27001 and SOC 2 are "in progress"; no report has been issued, so ask before relying on either.

Pricing is one credit per successful call on any of the five APIs, and failed calls cost nothing. The free plan gives 50 credits a month without a card. Starter costs EUR 24 a month for 500 credits (EUR 0.048 per call), Scale EUR 89 for 2,000 (EUR 0.0445), and one-time packs of 500, 2,000 and 10,000 credits cost EUR 29, 99 and 449. A full screening of one job against ten candidates costs about 12 credits: one for the criteria, ten parses, and one Rank call, or ten Match calls instead of Rank if per-candidate explanations are needed.

Evidence

The only published accuracy test is HireLayer's own benchmark of 7 October 2026: five fictional resumes written by HireLayer (French and English, two-column, Word tables, a rotated scan and an infographic), run through HireLayer's own dashboard and through sandbox or demo accounts HireLayer created at Affinda, Textkernel (with and without its LLM parser) and RChilli, with RChilli's resume indexing off.

HireLayer wrote the test resumes, chose the fields and ran the competitors' demos with default settings, so the result favors the fields HireLayer normalizes. The page says so, publishes the five files and the raw JSON, and reports the speed gap against itself. There is no independent benchmark, no named customer outcome with numbers, and the Wink figures (500+ companies, 70+ job boards, two minutes saved per application) are Wink's own.

Market position

HireLayer competes in the resume-parsing segment with Affinda, Textkernel, RChilli and HrFlow.ai, and with general extraction APIs such as Extend or DocuPipe that a team could point at a CV with its own schema. Its public price per CV is below Affinda's self-serve rate at low volume, and by HireLayer's own account the gap closes from roughly 10,000 CVs a month and Affinda's annual plans are cheaper at 50,000. Its publishing cadence is high: the blog shipped 14 comparison and how-to articles between 6 and 11 October 2026, about half of them with competitor prices. The vendor's benchmark page itself recommends running 50 to 100 of your own CVs through the free tier before deciding; the free tier covers that.

Resources

• Website and API documentation
• Pricing and integrations
• Resume parser benchmark and comparison with Affinda
• Wink case study
• MCP server and legal notice