MemContext closes the biggest gap in AI-assisted development: agents forget. Preferences, project context, and prior decisions vanish between sessions across Claude, Cursor, ChatGPT, and every other tool.
It's a persistent memory layer for MCP-compatible clients. Drop in an API key, add the MCP config, and your assistant starts remembering across sessions, tools, and projects.
• Hybrid Retrieval - Vector embeddings + PostgreSQL full-text search so both semantic queries and exact keywords surface the right memory • Evolving Memory - Temporal facts auto-expire and ranking adapts to feedback, so current truth always wins over stale context • Project-Scoped - Memories isolated per project so unrelated work never bleeds into the wrong session • Cross-Tool Sync - Works with Claude, Cursor, OpenCode, Codex CLI, Windsurf, ChatGPT, Gemini, and every MCP-compatible client • Turborepo Monorepo - Hono API, MCP server, Next.js dashboard, docs, and marketing site all in one workspace
I built the entire frontend and UI/UX for MemContext, both the marketing site and the product dashboard.
• Interactive SVG-based isometric hero showing AI tools syncing memory in real-time • Custom illustrations and product visualizations for each feature, hand-crafted in SVG • Animated micro-interactions powered by Framer Motion for hover, scroll, and state transitions • Data-driven dashboard UI with custom charts, stat cards, and live data tables • Cohesive dark + orange design system with consistent typography, spacing, and corner-bracket framing across every screen