When AI agents iterate on Figma designs, they often repeat the same mistakes — wrong spacing, mismatched colors, ignored brand guidelines — because they have no memory across sessions.
MCP Memory Gateway adds persistent feedback memory to any MCP agent:
claude mcp add memory -- npx -y rlhf-feedback-loop serve
The agent gets recall (check past mistakes before designing), capture_feedback (record what worked/failed), and prevention_rules (auto-generated from repeated failures).
Example: Agent sets wrong padding → user captures "down" feedback → next session agent recalls "NEVER use 8px padding on cards, use 16px" → follows the rule.
npm | GitHub
When AI agents iterate on Figma designs, they often repeat the same mistakes — wrong spacing, mismatched colors, ignored brand guidelines — because they have no memory across sessions.
MCP Memory Gateway adds persistent feedback memory to any MCP agent:
The agent gets
recall(check past mistakes before designing),capture_feedback(record what worked/failed), andprevention_rules(auto-generated from repeated failures).Example: Agent sets wrong padding → user captures "down" feedback → next session agent recalls "NEVER use 8px padding on cards, use 16px" → follows the rule.
npm | GitHub