Skip to content

Interface understanding as MCP tools

memi exposes repository audits, briefs, tokens, and scaffold plans through MCP so an agent can read the interface before it writes.

Public install target

Registry idio.github.memi-design/memi
Package@memi-design/cli@2.7.9
ModeWorks code-first with --no-figma; Figma context is optional.

Connect over Streamable HTTP

MCP SERVER URL
https://www.memoire.cv/api/mcp
  • memi_capabilities - discover remote and local execution boundaries.
  • prepare_design_agent_brief - generate a compact, executable design workflow.
  • get_design_ci_policy - create deterministic UI quality gates.
  • get_install_instructions - configure npm, MCP, Agent Skills, and agent kits.
  • get_public_proof - resolve canonical package, registry, docs, and proof links.

The remote server is stateless and scoped. Repository files, approved scaffolds, local audits, and Figma context stay in the npm MCP server so Memi never pretends Vercel has access to a caller's project.

Run the full engine locally

TERMINAL
npm i -g @memi-design/cli@2.7.9
memi mcp start --no-figma
MCP CONFIG
{
  "mcpServers": {
    "memoire": {
      "command": "memi",
      "args": ["mcp", "start", "--no-figma"]
    }
  }
}

v2.7.9 agent design tools

  • prepare_design_agent_brief - compact, standard, or full design-agent summaries.
  • scaffold_agent_design_files - dry-run or approved Atomic Design spec scaffolds.
  • run_audit - local design, UX, and craft evidence for a repo or route.
  • design_doc - agent-ready interface summaries from a live route.
  • get_tokens - token context without forcing Figma connection.