Skip to content

From v1 workbench to v2.5 agent design CI

Product Hunt rejected the relaunch when the public evidence did not show a large enough change. v2.5 fixes that by leading with the real upgrade: Memi is now a CLI, MCP server, Codex plugin, Claude skill, and design CI layer for agents that edit UI.

What changed since the previous launch?

The jump from v1 to v2.5 is the product change. The original v1 launch was centered on a macOS design workbench. v2.5 rebuilds Memi around agent design CI: Codex, Claude, Cursor, Grok, and MCP clients can now request compact design briefs, run token and accessibility checks, and create Atomic Design scaffolds from dry-run JSON before touching product files.

The package also moved from an app-only story to public distribution surfaces: npm package, GitHub Action, MCP Registry target, Codex plugin, Claude plugin manifest, shadcn examples, and installable agent skills.

Growth target and proof

DownloadsPre-growth baseline: about 750/week and 2,049/month before the v2.5 public push.
Target10x weekly downloads means roughly 7,500+ weekly installs after proof surfaces and directory listings are live.
Product HuntThe relaunch should wait until npm, MCP, Codex, Claude, docs, and this site all agree on v2.5.0.

Paste-ready short copy

PRODUCT HUNT
Memi v2.5 is agent design CI for AI coding agents. Codex, Claude, Cursor, Grok, and MCP clients can now understand a repo's interface before editing it: compact design briefs, token and accessibility audits, shadcn/Tailwind context, Codex and Claude skills, MCP tools, and spec-first UI scaffolds with dry-run JSON before writes.