Project Memory
Save durable project context so Copilot, repair flows, publish review, frontend generation, and MCP agents stay aligned.
Project Memory
Project Memory stores durable context for one generated project. Use it for decisions, constraints, preferences, warnings, and facts that AI should remember across future Copilot runs, repairs, publish reviews, frontend generation, and MCP agent work.
What To Save
- Business rules: "Invoices must be immutable after payment." - Data constraints: "Never expose password digest fields to AI data access." - Architecture decisions: "Use Stripe Checkout, not Payment Intents, for hosted checkout." - Frontend expectations: "The client portal is mobile-first and uses email login." - Deployment notes: "Production deploys through the configured Render hook."
How It Works
Add memory
Open Settings > Project Memory and write the fact, decision, constraint, preference, or warning.
Set scope
Choose a scope such as project, auth, frontend, deployment, data, agent, or resource.
Approve or pin important items
Approved active memories can be used by AI. Pinned memories stay easy to find for critical context.
Archive old context
Archive memory that is no longer true. Archived items remain available for audit but are ignored by AI.
Memory should be true and useful
Do not store secrets in Project Memory. Put API keys and private values in Environment variables or addon settings.
Why Users Care
AI gets better when it has stable product context. Project Memory prevents the same explanations from being repeated and helps different agents respect the same project rules.
AI Copilot /features/ai-assistant Use memory as context for generation and Ask mode MCP /agents/mcp Give authenticated agents the right project context Settings /settings Find Project Memory in the settings modal