Zaai Dev MCP server

Read from the package source without running it. The installed server may list more.

• add_reference

  Add a URL reference to a project's library \u2014 for when you've found a relevant page and want to save it. No screenshot required; stored with source_kind='mcp' so the UI distinguishes it from extension-pushed captures. Use when discovering inspiration, never to bulk-import URLs the user already has. Costs 5 credits.

• get_animation

  Returns just the animation data. Page \u2192 pageAnimations (keyframes vocabulary + detected motion libraries like Framer Motion, GSAP, AOS, anime.js, Lenis, Locomotive). Element \u2192 the element's CSS transitions, keyframes, hover/focus/active state diffs, and per-element library hints. Composite \u2192 per-element variants. Use this to give the LLM motion context without loading HTML and compu

• get_audience

  Returns only the audience slice: primary segment, optional secondary, needs / pain points, channels. Use when targeting copy or design at a specific audience. Costs 1 credit.

• get_brand_brief

  Returns the full published brand brief for a project: positioning, values, voice, audience, design intent, and decisions. Use this when the LLM needs the whole picture before writing copy or designing a component. If the project has no brief yet, the response still validates \u2014 most fields are empty strings or empty arrays.

• get_brand_tokens

  Returns the brand's design tokens: colors (with optional role), fonts (role / family / weights), radius scale, shadow scale. Use this when generating CSS, Tailwind config, or component styles that should match the brand. Costs 1 credit.

• get_capture

  Fetch the full payload for one capture by id. Includes everything the extension extracted (HTML, computed CSS, palette, fonts, animation data, media inventory, ancestry, page sections, manual eyedropper picks) plus fresh-signed Supabase URLs for the screenshot and thumbnail. Use list_captures or search_captures first to discover ids. Costs 1 credit per call; the 404 path (unknown id) doesn't bill.

• get_decisions

  Returns the brand + design decisions log for a project: what was decided, why, who decided, when, and which brief field (if any) it concerns. Use this to avoid re-litigating settled questions. Costs 1 credit.

• get_design_intent

  Returns the visual descriptors and anti-descriptors that constrain design exploration, plus the inspiration_summary. Use this before generating layouts, palettes, or component styles \u2014 it's the guardrail set the brief encodes. Costs 1 credit.

• get_doc

  Read a doc as an agent-ready block list: text blocks as markdown, capture blocks as structured references with role / treatment / note and full capture metadata (palette, typography, vibe, inferred component) plus the captured `code` (element outerHTML + computed CSS + build tokens like bg/color/radius/padding/font/shadow/border) so you build from real values, not a screenshot guess. Includes the

• get_html

  Returns just the HTML for a capture. Page \u2192 full page HTML; element \u2192 outerHTML of the picked element; composite \u2192 concatenated outerHTMLs with `<!-- Element N: selector -->` markers. Use this for component-extraction prompts ('rebuild this in Tailwind'). Costs 1 credit. Page captures can be large \u2014 budget your context accordingly.

• get_media

  Returns just the rich-media inventory for element + composite captures: videos (src/poster/autoplay/loop), images (src/srcset/alt), background-image URLs across the subtree, and carousel containers (Swiper, Glide, Splide, Slick, Flickity, Embla, `[data-carousel]`) with slide counts. Page captures return media:null (page-level palette/fonts only). Costs 1 credit.

• get_palette

  Returns just the colour palette for a capture. Page captures carry a 6-colour extracted palette (`palette`); element + composite captures carry the user's eyedropper picks (`manualPalette`). Use this when the LLM needs colour context without loading the full payload. Costs 1 credit.

• get_project_bundle

  Fetch the whole project as one agent-ready bundle: every doc's block list (text + roled capture references + approval state) plus the project's capture library. Use this to load full project context up front in a single call. Can be large. Costs 1 credit.

• get_references

  Paginated list of references (captures) for one project. Use this when the LLM needs to know what visual / interaction references the user has saved for a specific project \u2014 different from list_captures which is org-wide. Costs 1 credit per call.

• get_structure

  Returns just the Layer-Explorer teardown \u2014 the 'structure map' (BUILD-PLAN \xA76.6): a flat list of the captured subtree's nodes, each with depth, a root-relative rect, short selector, inferred component, headline computed style (display/background/color/radius/font), and flags (flex/grid, animated, brandToken, a11yIssue, hidden). Page \u2192 main-content subtree; element \u2192 the captured

• get_voice

  Returns only the voice slice: one-liner, tone descriptors, do-say / don't-say lists, example phrases. Lighter than the full brief when the LLM is only writing copy. Costs 1 credit.

• health

  Returns the Zaai Dev MCP server's status, version, and uptime. Use this to confirm the server is running before relying on auth-gated tools. No workspace access required. Useful as a smoke test from the MCP Inspector before configuring credentials.

• list_briefs

  List every brief on a project \u2014 id, type, name, status, and current-version metadata. Call this first when working with a project so you know what briefs exist; then pass `brief_id` into the targeted read tools (get_brand_brief / get_voice / get_audience / get_design_intent / get_brand_tokens) to target a non-default brief. Returns oldest-first by created_at. Costs 1 credit.

• list_captures

  List the user's Zaai Dev captures (newest first, paginated). Use this to discover what's in the library before calling get_capture for specifics. Each row carries id, type, title, URL, capturedAt, tags, and a thumbnail URL \u2014 but NOT the full HTML, computed style, or animation data. Use get_capture for those. Costs 1 credit per call.

• list_docs

  List every doc on a project \u2014 id, type, name, status, and current-version metadata. Friendly alias of list_briefs. Call first, then pass `doc_id` into get_doc. Returns oldest-first. Costs 1 credit.

• log_decision

  Append a decision to a project's decisions log: what was decided, optionally why, optionally which brief slice it concerns (e.g. 'voice.tone'). Use this when you make or recommend a brand- or design-relevant choice that future you should trace back. Costs 1 credit.

• search_captures

  Find captures matching a text query (case-insensitive substring match against source_title, source_url, and note). Returns the same summary shape as list_captures. Use this when you have a specific concept in mind ('hero', 'pricing table', 'stripe'); use list_captures to browse instead. Costs 1 credit per call.

• search_references

  Keyword search over a project's references (title / url / note in v1; pgvector ranking in v1.5). Returns scored matches with matched_field signals. Use when the LLM has a specific concept in mind ('hero', 'pricing table'). Costs 1 credit.

• whoami

  Returns the Zaai Dev workspace account this server is bound to: user id, organisation id, project scope, and current credit balance. Use this to confirm the configured ZAAI_API_TOKEN is valid before calling any other workspace tool. Does NOT debit credits \u2014 the workspace treats /me as a heartbeat, not user-facing work.

• Code scan3 source files scanned25/25
• –Live reliabilityno gateway calls yet and no remote to proben/a
• –Tool poisoningtools not inspected (local package is not executed); not countedn/a
• Auth qualitystatic API keys via environment variables6/15
• Maintenancerepository not readable: repo not found3/15
• Maintainer identityno repository or website to verify2/10

Safety score

55/100

higher than 10% of 1,337 Media servers

Tools

24

the typical Media server has 7

Tools that change things

0%

typical: 4%. Fewer write tools means less an agent can break

Downloads a week

32

more than 16% of 365 Media servers with a package

Zaai Dev: common questions

Is Zaai Dev MCP server safe?

With care: it is graded C, so read the findings first (55/100). Read the Zaai Dev safety report

How do I install Zaai Dev?

It runs on your machine. Copy the Claude Code, Cursor, VS Code or Claude Desktop config from the install section.

Does Zaai Dev need an API key?

Yes. The registry entry asks for ZAAI_API_TOKEN.

Is Zaai Dev maintained?

The latest release is v0.6.2.

What can I use instead of Zaai Dev?

Servers from other publishers that do the same job: Designer Skill MCP server, Uizze MCP server and Dembrandt MCP server.

How does Zaai Dev compare with other Media MCP servers?

safety score 55/100 (higher than 10% of 1,337 Media servers); tools 24 (the typical Media server has 7); tools that change things 0% (typical: 4%. Fewer write tools means less an agent can break). Full comparison.