Job search
RenderCV Review: YAML Resumes, the CLI, and the New Web App
Review RenderCV YAML and Typst CLI exports, an actual sample layout repair, and the web app’s AI data handling, pricing displays and untested limits.
RenderCV is useful to investigate when you want structured resume content in YAML and a repeatable local export. Assess the open-source CLI and the newer web product as separate workflows; the website advertises AI editing, account features and paid access. Keep the actual version, source and generated file together. Neither schema validation nor the product's layout and ATS assurances establishes that every employer will read your attachment correctly.
LandOffer publishes this review and offers job-search tools, giving us a commercial interest in this category. On October 14, 2026, we read official sources, inspected public web pricing and FAQ descriptions, and used an isolated local installation of RenderCV 2.8 with original fictional content. We did not enter the guest editor, sign in, upload a CV, use web AI or subscribe. Astrid and Bastian below are teaching cases. Compare supported materials with one recent role on LandOffer before choosing a tool.

Distinguish the current CLI from the website
The official repository describes an MIT-licensed resume builder with text-based input. Current CLI documentation uses YAML input and generates Typst source alongside PDF and other outputs. An older description calling the current rendering path simply a LaTeX resume generator would miss that distinction.
The current website advertises a form or YAML editor, existing-CV import, an AI career assistant and language support. Its paid offer lists PDF downloads, cover-letter generation, public CV links, job tailoring and GitHub synchronization. We inspected the public descriptions, not those private workflows or their output quality.
Choose by the task you need. Keeping local structured content and rendering a file is different from importing a document into a website, asking AI to edit it, maintaining an account or publishing a shareable link. Open-source licensing for the CLI does not establish that all web capabilities are free or locally processed.
You can use the CLI without making a web subscription part of the exercise. Conversely, a person wanting an interactive assistant may need to evaluate web features and terms that a local build cannot test. Keep those evaluations separate so a successful command does not become evidence for an untested website feature.
Understand the YAML learning and maintenance work
YAML gives you named fields and structured entries. Content, design and settings can be changed deliberately rather than rearranging every line by hand. The trade-off is learning the expected fields, indentation and permitted values, then inspecting the resulting file.
The CLI reference documents creating a starting file, rendering it and selecting output options. Our isolated environment reported version 2.8, Python 3.12.14 and Typst package 0.15.0. The version record identifies this exercise’s environment; check your own installation before reproducing it.
The generated starting file contained demonstration employers, credentials and metrics. We preserved it as evidence and replaced the CV content with original fictional facts before rendering. A renamed file is insufficient: read every remaining entry and remove all demonstration experience you cannot claim.
Keep the actual work facts and dates beside the YAML, so each entry can be checked. Schema-valid fields can still contain a wrong degree, employer, date or unsupported result. Check each career statement against its factual record, then validate the input structure separately.
Astrid completes an original local build
Astrid Lorne is a fictional junior data engineer. Her exercise record contains Python ingestion checks at fictional Lake Orchard Lab from June 2024 to August 2026, an associate degree completed in May 2024 and a separate practice sensor importer from September 2026.
Her original employer-work sentence is: “Checked Python ingestion batches against source row counts and recorded rejected identifiers for the data reviewer.” It identifies the comparison and reviewer boundary without claiming that she designed every ingestion service or achieved an invented reliability percentage.
The practice importer checks timestamps and writes accepted samples to SQLite. Invalid samples go to a separate log. We put that work under Projects, keeping it separate from employer production work. No production reliability metric was measured in the exercise.
A compact part of the original input looks like this:
cv:
name: Astrid Lorne
headline: >-
Fictional teaching sample
sections:
skills:
- label: Languages
details: Python, SQL
design:
theme: classic
entries:
degree_width: 3.4cm
This excerpt is not the entire file: the completed source also includes the experience, projects, education, location, reference and settings fields. It is original teaching content, not a candidate's uploaded resume or a vendor-created assessment result.
We rendered the full YAML locally. The successful command created a PDF, Typst source, Markdown, HTML and a page PNG. No personal information or candidate document was submitted to the web editor. Installation downloads and local file generation are separate from the web service's account and AI processing.
Inspect the outputs and repair an actual layout issue
The final PDF contained one US Letter page and was 29,338 bytes. The other generated files were a 4,400-byte Typst source, 891-byte Markdown file, 3,338-byte HTML file and 129,899-byte PNG. These are measurements for one short fictional sample, not general size or speed promises.
The first Classic-theme proof placed Associate degree in a narrow column and wrapped it across five lines. We preserved that output, adjusted the degree-column width and rebuilt. An intermediate 2.3-centimeter width still split the word degree; the final 3.4-centimeter setting produced a readable education row without changing the credential.
The design guide describes overriding theme settings. We also inspected the installed 2.8 schema for design.entries.degree_width, then used that supported field in the actual source. The correction concerns this layout and text, rather than every theme or every degree name.

| Actual local check | Completed result | Boundary retained |
|---|---|---|
| Replace generated demonstration content | Original internship, practice project and credential inserted | Fictional facts; no real applicant history |
| Render the full input | Five output types produced | No DOCX output produced by this command |
| Inspect the education row | Degree-column width changed to 3.4 centimeters | A local content/layout repair, not a tested web-editor behavior |
| Inspect final page | Readable headings and bullets; no observed clipping or overlap | One short sample, not a universal layout guarantee |
| Extract PDF text | Sections remain Experience, Projects, Education, Skills, Reference | Decorative location glyph and line-break artifacts remain in raw extraction |
| Inspect the HTTPS annotation | Python public reference target retained | Reference is not Astrid's portfolio; no private-access test |
The raw extracted text contained a decorative location symbol and hyphenation artifacts at wrapped words. We kept that output with the PDF instead of equating a successful build with every parser receiving identical text. The generated Markdown and HTML are additional artifacts; neither is automatically the employer's requested attachment.
This observed layout repair also gives the website's strong formatting language a practical boundary. The valid sample still required a degree-column adjustment after its first page inspection. We did not test whether the web editor imposes different constraints or prevents this particular arrangement.
Validation catches syntax without verifying your career
To examine one validation boundary, we copied the input and changed the experience end month to the impossible value 2026-13. RenderCV rejected the separate input with an error identifying cv.sections.experience.0.end_date. The valid final source and its exports remained available.
That is a useful, narrow observation. It shows a date-format check on this installed version. It does not verify that Astrid worked at an employer, earned the credential or performed the listed task. A plausible but inaccurate date can remain structurally valid.
Use an error to locate the affected field, then correct it from your own record. Do not guess a month merely to make validation pass. If the underlying date is uncertain, reconstruct it through appropriate personal records before writing a precise statement.
Keep accepted revisions in the YAML and rebuild the requested outputs. Editing only the generated PDF would leave the next render carrying the old content. Save the source version and output together so you know which facts the application copy contains.
Read web pricing and limits before committing
The native public pricing view showed a selected three-month Recruit option at $10 per month and a separate line saying $31 billed every three months with automatic renewal. We retained both displays. Multiplying the headline rate by three would give $30; it would not reproduce the displayed billed amount. We did not open checkout to establish the final charge or explain that difference.
The February 25, 2026 terms describe free and paid plans, a free storage limit of up to six CVs, recurring billing, cancellation through the site or Polar portal, and non-refundable fees. Their cycle description mentions monthly or annual billing, while the current public selection offered a three-month term. Keep that documentation difference unresolved until the actual purchase conditions are clear.
An advertised guest entry point does not establish unlimited ongoing downloads, AI access or persistent storage. We did not enter it. The public continuation area included agreement language, and no account or service continuation was completed during this review.
Decide which missing capability would justify paying. If the local CLI already produces the needed reviewed PDF, web AI or public links are separate optional tasks. If you want those tasks, verify their limits, data handling and actual billed period before making a commitment.
Bastian completes a local versus web decision
Bastian is a fictional analyst with one accurate source resume and a three-month job-search plan. He needs a current PDF now, may want feedback on supported SQL wording later, and has no present need for a public resume link or GitHub account synchronization.
He completes the following decision record using the observed build and public descriptions. It is an original exercise, not a claim that he purchased, subscribed or received AI edits.
| Bastian's task | Completed choice | Evidence or unresolved issue |
|---|---|---|
| Create a reviewed PDF from factual content | Use the local YAML/CLI route first | The actual sample build and file checks demonstrate that route |
| Obtain suggestions for a supported SQL sentence | Defer web AI until its processing terms fit his choice | No generated rewrite or quality result inspected |
| Store an editable source | Keep a local backup with the version record | Web guest persistence remains documentation-only |
| Publish a public CV link | Defer because the audience is not needed | No link created, discoverability or access control tested |
| Consider the selected paid term | Record $31 per three months, subject to checkout verification | Headline and billed display differ; no charge authorized |
Dividing the displayed $31 by three gives roughly $10.33 per month as simple budgeting arithmetic. His budget remains provisional because checkout was not inspected; the calculation does not establish a refund or value for him. His completed decision is to retain the local build and postpone the paid web tasks he does not currently need.
His supported editing sentence remains: “Compared SQL output with source records and documented unmatched identifiers for review.” An assistant's suggestion would still need to preserve that contribution and avoid adding measured business results he cannot substantiate. No web assistant produced this wording in the exercise.
Match data handling to the features you use
The privacy policy, revised February 25, 2026, describes account CV storage and browser-memory guest content. It separately says AI editing and PDF import send content to Google's Gemini models through Cloudflare's AI Gateway for processing, while stating that CV content is not used to train AI models. These are published practices, not a network audit or independent provider verification.
The public FAQ described exporting a ZIP with YAML, PDFs and Typst source. We did not perform that export, deletion or GitHub sync. Keep your own backup before relying on a later account export, and distinguish processing permission from the decision to publish a public CV link.
Choose one recent role on LandOffer, keep supported facts in an identifiable source revision and inspect its actual export. Identify the web task you need before accepting its data processing or billing conditions. Confirm the employer's requested format and visible imported fields when you later complete the separate application process.


