Overview
Taste Lab and memi are both open-source tools that bring AI into the design workflow, but they serve very different purposes. Taste Lab is a focused CLI tool that reverse-engineers any website's design taste into structured files for AI agents. memi is a full macOS workbench where AI agents run with project memory, design context, and Figma integration.
Feature Comparison
| Feature | Taste Lab | memi |
|---|---|---|
| Core Purpose | Reverse-engineer any website's design taste into a structured file for AI agents. | A macOS workbench for running AI agents with project memory, design context, and Figma integration. |
| Target User | Developers and designers who want to replicate or learn from existing website designs using AI. | Product designers and teams who need a local, persistent environment for AI-assisted design work. |
| Output Format | Markdown (.md) and JSON (.json) files containing Design Map (tokens) and Taste DNA (principles). | Project memory in Markdown/YAML, run receipts, FigJam-ready boards, and design artifacts. |
| AI Integration | Works with Claude Code, Gemini CLI, Cursor, Windsurf, GitHub Copilot, Bolt, v0, Lovable, and more via skill files. | Native integration with Codex and Claude Code; also supports OpenCode and Figma plugins. |
| Design Analysis Depth | Four-step pipeline: Extract measurements, Detect patterns, Infer taste, Observe output. Produces 4 taste principles with trade-offs. | Provides UX audit skills (tenets & traps), design memory, and Figma token/component inspection. |
| Platform | CLI-based, works on any OS with Node.js and Playwright. | macOS app (signed) with a Studio UI; also offers a CLI via npm. |
| Figma Integration | None directly; output can be used in Figma via taste-tokens.css. | Deep Figma bridge: pull tokens, components, trees, screenshots; live plugin connection for audits. |
| Memory / Persistence | Output files are static; no built-in project memory beyond the generated files. | Design memory layer: editable project state with decisions, tokens, research, and notes that agents can read and reuse. |
| Setup Complexity | Requires cloning a repo and installing Playwright MCP; relatively simple for CLI users. | Download and install the macOS app; configure agents and Figma plugin; more setup for full workflow. |
| Pricing | Open-source (free); no pricing information on the website. | Open-source (free); no pricing information on the website. |
Pricing
Both Taste Lab and memi are open-source and free to use. There are no paid tiers, subscriptions, or premium features mentioned on their websites. This makes them accessible to individual designers and teams alike.
Pros and Cons
Taste Lab
Pros:
- Deep, structured design analysis with explicit trade-offs
- Works with many AI tools and IDEs via skill files
- Output is immediately usable in AI agents for building
- Four-agent pipeline ensures quality and reduces slop
Cons:
- No built-in project memory or persistence
- No Figma or visual design tool integration
- Requires manual setup of Playwright and MCP
- Static output; no iterative workflow support
memi
Pros:
- Persistent design memory that agents can read and reuse
- Deep Figma integration with live token and component pull
- UX audit skills for design quality checks
- Signed macOS app with a clean UI for managing runs
Cons:
- macOS only; no cross-platform support
- More complex setup for full workflow
- Figma integration requires plugin installation
- Heavier tool; may be overkill for simple design analysis
Verdict
Choose Taste Lab if you need a quick, structured design breakdown of any website for your AI agent, especially if you work across multiple platforms. Choose memi if you're a product designer on macOS who needs a persistent, Figma-integrated workbench for ongoing AI-assisted design projects with memory and audit capabilities.
memi
Taste Lab