MotionLint
Frameworks & Eval · tested 2026-08-09 · re-test due 2026-11-09 · by the Hlido desk, not the vendor
In short: Reviews the animation your coding agent cannot see, and splits the job honestly — a deterministic linter for what is measurable, a vision model only for what is not.
5 PASS · 2 FAIL of 7 public-surface claims
Quick answer
MotionLint scores 80/100 (STEADY) on Hlido’s independent, hands-on test (reviewed 2026-08-09). STEADY (80) for a well-drawn problem (agents cannot see what renders), a design that keeps the measurable half deterministic and key-free so it can gate CI, findings that cite the exact encoded rule and value, accessibil Pricing: Open source (free entry point documented).
MotionLint's framing is the sharpest thing about it: coding agents read JSX and CSS and are structurally blind to what a user actually watches happen, so it captures the running app instead of the source. The implementation respects a distinction most tools in this space blur. `motionlint audit` measures every animation in the browser over CDP and lints duration, easing, stagger and exits against encoded numeric standards — no API key, no LLM, SARIF output and exit codes for CI. `motionlint review` is where the vision model goes, returning ranked findings across twelve UX dimensions. Keeping the deterministic half deterministic is what makes this usable as a CI gate rather than a suggestion box, and every finding cites the exact rule and value it was measured against (≤300ms for UI motion, 200–500ms for modals, ease-out on entrances, a prefers-reduced-motion fallback) rather than asserting taste. Accessibility is treated as a lintable rule, not a footnote. A native MCP server puts findings next to the code being written, and `motionlint tune` exports a diff an agent can apply — closing the loop rather than stopping at diagnosis. The claim to treat carefully is '100% recall, 0% false positives on a 24-fixture stress test': a 24-fixture suite is a small, self-authored sample, and a perfect score on it says considerably less than the phrasing implies. There is no third-party validation, no version and no adoption evidence on the captured surface.
Why STEADY
STEADY (80) for a well-drawn problem (agents cannot see what renders), a design that keeps the measurable half deterministic and key-free so it can gate CI, findings that cite the exact encoded rule and value, accessibility as a first-class lint, and an MCP path plus an exportable diff that closes the loop back into the agent. Held below VITAL because the only quality evidence offered is a self-authored 24-fixture suite reported as 100% recall / 0% false positives — too small a sample to carry that phrasing — with no version, adoption or independent validation on the surface.
Public-surface checklist
- PASS Homepage loads (required)
- PASS Primary value prop (required) — 'Every animation needs a reason' — reviews design and motion of the running app
- PASS Cta present (required) — `npm install -g motionlint` with copy button and GitHub link
- PASS Pricing or access — Open source, MIT, live on npm
- PASS Docs present (required) — Features, Standards and worked audit output on the page
- FAIL Quantified claim supported — 100% recall / 0% false positives rests on a 24-fixture self-authored suite
- FAIL Third party validation — None published
What we saw
2 screenshots captured by the Hlido engine during the reviewed run (run-9662f720843d8c21-www-resila-ai). Our own captures — not vendor marketing material.
What it does well
- Captures the running app rather than the source, which is the actual blind spot for coding agents
- Splits deterministic measurement (audit, no LLM, no API key) from judgement (review, vision model) instead of blending them
- SARIF output and exit codes make the deterministic half a usable CI gate
- Every finding cites the exact rule and measured value — ≤300ms UI budget, 200–500ms modals, ease-out entrances
- Treats prefers-reduced-motion as a lintable accessibility rule, not an afterthought
- Native MCP server for Claude Code, Cursor and other MCP-aware clients
- `motionlint tune` exports a diff the agent can apply — diagnosis through to fix
- MIT-licensed and published on npm
What it fails at
- '100% recall, 0% false positives' rests on a self-authored 24-fixture suite — too small to support that phrasing
- No independent or third-party validation on the captured surface
- No version, changelog or adoption evidence published
- The `review` path depends on a vision LLM, so its findings are neither free nor reproducible the way `audit` is
- Not tested hands-on by Hlido
Red flags
- 'Validated at 100% recall, 0% false positives' is stated without qualification, but the sample is a self-authored 24-fixture stress test. A perfect score on 24 self-chosen fixtures is weak evidence presented in strong language.
Best for
- Teams whose UI is largely written by coding agents and who need motion quality checked by something that can see it
- CI pipelines wanting a deterministic, key-free animation gate with SARIF output
- Developers who want motion accessibility (prefers-reduced-motion) enforced rather than remembered
Not recommended for
- Teams needing validated, independently-benchmarked review quality before adoption
- Design systems with deliberately unconventional motion that would fight the encoded budgets
- Anyone expecting the LLM `review` path to be as reproducible as the deterministic `audit` path
Pricing & access
- ModelOpen source
- Free entry pointYes — a free tier or open-source edition is documented
- Pricing findable on the public surfacePASS Open source, MIT, live on npm (tested 2026-08-09)
Derived from Hlido-held evidence only (engine checklist + editorial text); quotes are verbatim from the scorecard; not vendor-supplied; re-derived daily. Verify current prices on the vendor's pricing page. Last verified 2026-08-09.
Related agents
Agent relevance
CLI MCP Behavioral-testable
Agentic-Commerce Readiness 62/100 · INTEGRABLE
Independent readiness for agent delegation & transaction. How it’s scored · check live
Built for the coding-agent loop: a native MCP server surfaces findings inside Claude Code or Cursor, the CLI runs in CI with SARIF and exit codes, and `tune` exports a diff the agent applies. The stated purpose is to give an agent eyes it does not have.
Agent-friendly score: 9/10
Score over time
The longitudinal record — every point is the score as published on that date. Raw series.
Evidence
- Audits the running web app over CDP rather than reading source — source (2026-08-09) verified
- `motionlint audit` is deterministic — no API key, no LLM — with SARIF output and CI exit codes — source (2026-08-09) verified
- `motionlint review` uses a vision LLM for ranked findings across 12 UX dimensions — source (2026-08-09) verified
- Findings cite encoded standards — ≤300ms UI motion, 200–500ms modals, ease-out on entrances, prefers-reduced-motion fallback — source (2026-08-09) verified
- Native MCP server for Claude Code, Cursor and other MCP-aware clients — source (2026-08-09) verified
- MIT-licensed and live on npm via `npm install -g motionlint` — source (2026-08-09) verified
- '100% recall, 0% false positives' generalises beyond the 24-fixture self-authored suite — source (2026-08-09)
- Independent validation, version or adoption evidence — source (2026-08-09)
scorecard.json · transparency passport · registry · methodology
Embed this trust badge
Markdown
[](https://hlido.eu/check/?agent=bobaba99-motionlint)HTML
<a href="https://hlido.eu/check/?agent=bobaba99-motionlint"><img src="https://hlido.eu/badge/bobaba99-motionlint.svg" alt="Hlido trust score"></a>