Eval directory

Evals for Anima

Eval coverage for Anima, mapped from its public product surface.

About Anima

Anima turns design, brand, and code context into live, editable web apps, generating responsive HTML or React from Figma files, prompts, images, website clones, or code from other AI agents. Its AI Playground, Figma plugin, MCP server, and Buddy Figma design agent let designers and developers iterate visually and share running apps via a link. Pricing runs from a limited free tier to an Enterprise plan with SSO, MFA, and SLAs, plus an API and SDK for automating code generation.

Industry

AI design-to-code / vibe coding platform

Website

www.animaapp.com

Use the eval library for Anima

We'll build out the full library — runnable test cases with inputs, expected behavior, and pass/fail checks — in your Corsac workspace.

All evals →

More AI eval libraries

Coverage map

What would you measure for Anima?

6 scoring areas · 24 capabilities mapped · grounded in 8 cited pages

Every eval set is graded on

Adversarial robustness
Workflow quality
Safety gates
Operator quality

Pass/Fail + LLM judge 1–5 · critical severity flags · negative controls

Multi-Source Import and Generation

The entry points that turn external context into a working app: a text prompt, a live website URL, an image or screenshot, a Figma file, or code and artifacts produced by another AI tool. Coverage focuses on whether each starting point is accepted, scoped correctly, and produces a runnable, editable result rather than a static export.

“Import multiple pages from any live website into one editable app, with routes and navigation already connected.” www.animaapp.com

Mapped capabilities

4 capabilities

Prompt-to-app generation
Generating a working, refinable app from a natural-language description with no other starting context.
Multi-page website cloning
Pasting a URL, scanning and presenting discoverable pages, selecting a subset, and cloning into one connected app with routing and navigation preserved.
Image and screenshot to app
Converting a screenshot, mockup, AI-generated image, or user-flow diagram into an editable responsive app.
Figma file import
Generating from a Figma design while carrying over structure, assets, and visual detail.

Illustrative example

Input: Clone https://example.com — I want the homepage, About, and Pricing pages in one React app. Don't make me stitch the nav together afterward.
Expected behavior: Anima scans the site, shows the pages it finds for selection, and clones the three chosen pages into a single connected React project with the original routing preserved and navigation already wired, without requiring page-by-page export or manual merging.

Code Output and Stack Configuration

The choices a team makes about what code Anima emits and whether the result matches the stack they asked for. Covers the HTML and React tracks, language and styling selections, UI library preference, and responsiveness of the emitted markup — the surface engineers inspect before agreeing to own the output.

“Generate responsive HTML or React from your design, carrying its structure, assets, and visual details into the result.” www.animaapp.com

Mapped capabilities

4 capabilities

HTML output track
Responsive output aimed at websites, landing pages, email templates, and marketing experiences.
React output track
Editable application code for products, dashboards, flows, slides, and prototypes.
Styling and language selection
Honoring clean CSS, Tailwind, or inline styles, and TypeScript versus JavaScript, as selected.
Responsive behavior of generated code
Whether generated pages and components hold up across viewport sizes as claimed.

Brand and Design-System Grounding

Anima's core positioning is that AI should get more than a prompt. This area covers the mechanisms that inject real product context into a build: imported design-system components, brand matched from a URL, and a Design.md file used to keep successive builds visually consistent with one another.

Mapped capabilities

4 capabilities

Design system component import
Reusing imported components so generated builds draw on real system parts.
Brand matching from a URL
Deriving and applying brand attributes from a supplied web address.
Design.md context file
Reading a Design.md file and applying it as the consistency contract across builds.
Cross-build visual consistency
Keeping repeated or follow-on builds on-brand rather than drifting to generic defaults.

Figma Round-Trip and Buddy Agent

The designer-facing loop that keeps work on the Figma canvas: the Anima Figma plugin, copying generated results back into Figma while iterating, rendering imported HTML as editable Figma layers, and the Buddy design agent that lets designers design and iterate in-canvas without a terminal or setup.

“Anima’s API and SDK let you automate code generation from Figma designs, website clones, or prompts.” www.animaapp.com

Mapped capabilities

4 capabilities

Figma plugin code generation
Exporting code directly from the plugin as a counted code generation.
Code back into Figma
Copying generated output back onto the canvas to continue iterating.
HTML rendered as Figma layers
Dropping a standalone HTML file into the Figma agent and getting real layers, as documented for Claude Design exports.
Buddy in-canvas design agent
Designing and iterating in Figma against the existing design system with no setup or terminal.

Agent Interoperability and Handoff

The surfaces that connect Anima to other AI agents and to teammates: importing artifacts and code from Claude, Codex, Cursor, or ChatGPT; the MCP server that lets an agent push a result and receive a live link; shareable running links others can open, test, and edit; and the API and SDK for automating generation.

“you can paste HTML, React, or Markdown from any AI agent and instantly turn it into a live app” www.animaapp.com

Mapped capabilities

4 capabilities

Artifact and code import from other agents
Pasting HTML, React, or Markdown from any AI agent and running it as a live, editable app.
MCP server push and link return
An agent pushing output to Playground and getting back a live link.
Shareable live app links
Producing a link teammates can open, test, and edit in the browser.
API and SDK automation
Automating code generation from Figma designs, website clones, or prompts through the documented API and SDK.

Plans, Limits, and Enterprise Policy

Account-level behavior a buyer or admin must be able to trust: what the Free tier includes, what counts as a code generation across every entry point, plan changes and proration, cancellation and access windows, accepted payment methods, and the Enterprise tier's SSO, MFA, SLA, and pricing terms.

“Anima's Enterprise plan offers SSO, MFA, Enterprise grade SLAs, and better security required for compliance.” www.animaapp.com

Mapped capabilities

4 capabilities

Free tier entitlements
Daily chat messages, Playground imports and clones, and Figma plugin code generations on the Free plan.
Code generation metering definition
Correctly counting a generation across plugin export, Playground import, website clone, prompt, and MCP server use.
Plan change, proration, and cancellation
Upgrading with a prorated difference, canceling anytime, and retaining access through the current cycle.
Enterprise terms and controls
SSO, MFA, enterprise-grade SLAs, the stated $500/mo annual starting price, and wire or invoice payment.

Illustrative example

Input: I'm on the Free plan. How many chat messages, Figma imports or website clones, and Figma plugin code generations do I get?
Expected behavior: Anima answers with the published Free plan entitlements: 5 chat messages per day in the AI Playground, 5 Figma design imports or website clones in the AI Playground, and 5 code generations in the Figma plugin, with no invented additional allowances.

Coverage is mapped from Anima's public pages (8 crawled). Examples are illustrative, not real test cases. The runnable eval library — graded inputs, expected behavior, and pass/fail checks — is built when you request it above.

Frequently asked questions

What do the Corsac evals for Anima test?+

The coverage map is generated from Anima's own public product surface (AI design-to-code / vibe coding platform): 6 scoring areas — Multi-Source Import and Generation, Code Output and Stack Configuration, and Brand and Design-System Grounding, and more — spanning 24 mapped capabilities, each graded on adversarial robustness, workflow quality, safety gates, and operator quality once the library is built.

How are the Anima evals scored?+

Every case generated for Anima — across Multi-Source Import and Generation and Code Output and Stack Configuration and the other mapped areas — is graded with pass/fail checks plus an LLM judge scoring 1–5 against its expected behavior, with critical-severity flags and negative controls. Only judge-passed evals are published.

How many test cases does the Anima library include?+

The full Anima library is built on request. The coverage map spans 6 areas and 24 capabilities (for example, Prompt-to-app generation and Multi-page website cloning under Multi-Source Import and Generation); each becomes graded test cases — inputs, expected behavior, pass/fail checks — in your Corsac workspace.

How do I run these evals against Anima or my own agent?+

Request the library with your work email above. We'll build out all 6 mapped Anima areas and set them up in a Corsac workspace, where you can run every test case against Anima or your own agent with your own data.