Arcumet Review: strong biology tool, clear evidence (80/100)

Arcumet is a free, open-source Claude Code tool for interactive, sourced biology figures.

Fact	Detail
Domain	anima.arcumet.com
Category	Interactive, sourced biology figures for Claude Code
Pricing	Free; open source under the MIT license
Crawl stats	15 pages crawled on 2026-10-19

Arcumet earns 80/100 because it makes its job clear and its first-use path unusually straightforward. The product is a Claude Code skill for generating interactive, sourced, isometric biology figures, with real structures, a toggleable mechanism, and citations for every number on screen.

The strongest results are usability at 100/100, accessibility at 98/100, and writing quality at 92/100. The site explains the product directly, gives visitors live examples, and maintains a careful scientific voice. Technical SEO also scored 85/100, with all 15 crawled URLs returning 200.

The weaker areas are visual execution at 59/100 and docs and self-serve help at 59.5/100. Audience and messaging scored 70.7/100, while performance scored 62/100 because the available crawl did not provide complete lab or field measurements. The result is a focused open-source developer tool with a strong evidence standard and a clear opportunity to deepen presentation and support materials.

Arcumet makes its product legible within the first screen: it is a Claude Code skill that turns a biology idea into an interactive, sourced, isometric figure. The homepage combines that promise with three concrete boundaries—real structures, a toggleable mechanism, and a citation for every number on screen—and links six live figure examples. That is stronger than a generic visualization claim because visitors can test the work directly. The positioning gap is audience specificity: the supplied homepage does not say whether the primary user is a researcher, educator, student, or learner, and it contains no attributable testimonial or case study. Add one role qualifier and show a citation interaction inside the first figure so the evidence standard is experienced, not only stated.

The messaging explains the immediate job well: a Claude Code user can ask for a biology figure, see live examples, read the install command, and understand that the material is educational rather than medical advice. The evidence philosophy is unusually specific, separating established biology, one leading model, rodent work, and human work while citing numbers with species, conditions, and units. The remaining decision friction is audience and context. The page does not choose a primary role, and the supplied homepage contains no attributable testimonial, review score, or case study. It also gives no comparison for when an interactive sourced figure is preferable to a static diagram or manual explanation. Name the lead role and add a compact use-when panel.

Arcumet gives first-time visitors a direct path from comprehension to proof. The headline states what the product produces, the adjacent Open a live figure action supplies a clear next step, and the page keeps the GitHub source, README, license, install command, and named examples in the same journey. The mobile walkthrough preserves this hierarchy and stacks the examples without an observed blocker. One small clarity issue remains in the repeated figure controls: bare dark and light links rely on nearby card titles to identify their destination. Expand each label to include the figure name, such as a figure title followed by its theme, so the control remains understandable when read independently.

Arcumet has a strong accessible-content baseline in the sampled pages. The homepage reports a main landmark, one h1 followed by orderly h2 and h3 headings, seven images with non-empty alt text, no positive tabindex values, and lang="en". Sampled figure pages report zero missing image alt attributes and zero controls without a detected label or ARIA name. The verified issue is structural: the stimulants-dat, dopamine-receptors, and reward-addiction pages each begin with h2 and have no page-level h1. Add one concise visible h1 to the shared figure template. Lighthouse scores, computed contrast, and live keyboard behavior were not supplied, so this is a strong partial audit rather than a full conformance claim.

Arcumet presents a restrained visual system: a clear hero-to-gallery sequence, a dark primary action, square cards, and consistent black-and-white treatment across the supplied desktop and mobile screenshots. The main execution weakness is touch ergonomics. The 390px homepage reports 30 small tap targets, including top links measuring 38px high; sampled figure pages include a 112×29px Pause motion control and 13×13px native controls. Figure text is also compact, with 12px paragraphs, 12.5px links and buttons, and 13.3px inputs. Raise interactive hit areas to at least 44×44px with 8px separation, enlarge or wrap compact controls, and keep input text at least 16px. The page has no observed horizontal overflow on the homepage.

The crawl cannot establish a Core Web Vitals verdict: both mobile and desktop PageSpeed Insights records returned no lab or field metrics, and the sampled response headers contain no TTFB. The clearest available signal is payload size. The homepage transferred 4,625 KB across 14 requests, while sampled example pages transferred 29–33 KB. The homepage also delivers gallery thumbnails at substantially larger intrinsic dimensions than their displayed slots, including a 1,012×750 image in a 345×256 slot. Identify the asset behind the 4.6 MB transfer, compress or defer non-critical media, and provide right-sized responsive thumbnails. The homepage response was a cache HIT through Varnish/Fastly and used gzip, which are useful delivery foundations.

Arcumet’s writing is a major strength because it makes technical claims specific without turning them into sales language. The hero answers the product question directly, while the figure content pairs quantitative or mechanistic claims with peer-reviewed literature, named experimental models, and explicit distinctions between established biology, models, animal work, and human work. The estrogen example even states when human evidence is weak, preserving uncertainty instead of smoothing it away. This gives the site a consistent scientific voice: direct, sourced, and careful about what a figure can support. The practical improvement is to carry that same clarity into standalone page metadata and headings, so the editorial quality is as visible to search systems and assistive technology as it is to readers.

The current crawl shows no sitewide suppression pattern: all 15 crawled URLs returned 200, the HTTP root made one 301 hop to HTTPS, and raw and rendered content were effectively identical. The main resilience exposure is URL duplication. Seven dark/light figure pairs share duplicate-title signals, while sampled page facts report canonical:null; without a deliberate policy, search signals and crawl attention can split between presentation variants. Select one preferred URL per figure, canonicalize or noindex the alternate, and align internal links with that choice. The domain was registered on 2026-01-27, but Wayback returned no CDX rows and Google Search Console was not connected, so historical stability and actual traffic changes cannot be measured from this run.

Arcumet demonstrates strong editorial control over scientific claims. Quantitative and mechanistic statements are grounded in specific publications, and the figure copy distinguishes evidence from Drosophila protein, human DAT, animal models, and human studies instead of collapsing them into one conclusion. That discipline is reinforced by explicit uncertainty, including the statement that human evidence is weak where the underlying evidence is weak. The content QA gap is structural rather than scientific: all 14 figure embed pages lack meta descriptions and h1 headings, while light and dark variants duplicate title tags without canonical differentiation. Add a descriptive meta description and page heading to each figure, then point presentation variants to the preferred URL so the technical layer reflects the care of the prose.

Arcumet explains the product directly, but its self-serve support remains concentrated in one landing page and the linked README. A full API reference is not necessary for this visual tool, but Claude Code users would benefit from a short guide covering installation, figure controls, source trails, and common workflow questions. Machine-readable support is also incomplete: https://anima.arcumet.com/llms.txt returned 404. Add a focused usage guide or inline FAQ, then publish an /llms.txt file with the same factual boundaries as the public page.

Arcumet is broadly crawlable and server-rendered: all 15 crawled URLs returned 200, the HTTP root redirected once to HTTPS, and sampled raw and rendered word counts matched, including 395 vs 395 words on the homepage and 984 vs 984 on semaglutide pages. The main technical SEO issue is indexation ambiguity. Seven dark/light pairs carry duplicate-title signals, and sampled pages report canonical:null. Choose one indexable URL per figure, make the alternate policy explicit, and add self-referencing canonicals to intended indexable pages. Two neuron variants also exceed the 390px mobile viewport at 413px scroll width.

Arcumet is a strong fit for Claude Code users who want interactive, source-traceable biology explanations. Its clearest wins are the direct product promise, the unusually usable first-use experience, the strong accessibility baseline, and scientific writing that distinguishes animal models from human evidence.

The fixable weaknesses are concentrated in finish and support: design execution scored 59/100, docs and self-serve help scored 59.5/100, and performance scored 62/100 with incomplete measurement evidence. Deepening the visual system, adding dedicated guidance, and clarifying the remaining performance evidence would make the product easier to adopt without changing its focused scope.

For buyers who value free, open-source biology visualization inside Claude Code, the evidence supports a positive verdict.

This is a 13-dimension review based on a 15-page crawl dated 2026-10-19, the site's product facts and page content, and the public usability, accessibility, editorial, risk, and technical findings supplied for this review. SiteList combines those crawl observations with the stated dimension scores; it does not add claims beyond the supplied evidence.

Two dimensions were excluded as not applicable: 12 · Decision-support surfaces and 13 · Review-content integrity. Google Search Console was not connected or accessible. Performance enrichment also remains incomplete because the available inputs contained no lab or CrUX metrics and no TTFB value.

Read the SiteList methodology for the scoring framework and data notes. The byline describes this as a 13-dimension review; the crawl date is 2026-10-19.