Blog

Join Waitlist

Blog

Join Waitlist

Memory Store for Cursor

Memory Store for Cursor: Developer Guide

Turn Cursor into a context-aware development partner that remembers your codebase, decisions, and patterns across sessions.

What is Memory Store?

Memory Store transforms Cursor from a stateless AI into a persistent development partner. It remembers your business logic, architectural decisions, API contracts, infrastructure setup, and coding patterns—delivering contextually aware assistance when you need it.

Beyond Code: Helps with your entire delivery lifecycle—commits, infrastructure, CI/CD, releases, and deployments.

Without Memory Store:

  • Repeatedly explaining architecture and business logic

  • Losing context about APIs and data flow

  • Re-describing setup and dependencies

  • Generic suggestions that don't fit your patterns

With Memory Store:

  • AI knows your codebase structure and domain

  • Auto-retrieves relevant context for each task

  • Remembers architectural decisions and rationale

  • Tracks tech stack, dependencies, and setup

  • Learns team patterns and conventions

  • Cross-repository understanding

Installation

Step 1: Install MCP Extension

  1. Open Cursor → Settings → Extensions

  2. Search "MCP" (Model Context Protocol)

  3. Click Install

Step 2: Add Memory Store Server

  1. Cursor Settings → MCP Servers → Add Server

  2. Use this configuration:

{
  "mcpServers": {
    "memory-store": {
      "url": "https://beta.memory.store/mcp",
      "transport": { "type": "http" },
      "oauth": {
        "enabled": true,
        "discovery_url": "https://beta.memory.store/.well-known/oauth-protected-resource"
      }
    }
  }
}

Step 3: Authenticate

  1. Start a new chat: What do you know about this project?

  2. Browser opens → log in at beta.memory.store

  3. Approve connection

  4. Server shows as connected

Step 4: Verify

Ask: Can you check my memory store setup?

Note: No need for @memory-store tags—just talk naturally.

Step 5: Configure Cursor Rules (Critical)

  1. Settings (Cmd/Ctrl + ,) → Features → Rules for AI

  2. Paste:

How to Use

Works naturally—no special commands needed.

Record:

  • "Remember: our API uses OAuth2"

  • "Store this: we deploy via GitHub Actions"

  • "Database is PostgreSQL 15 on AWS RDS"

Recall:

  • "What's our authentication setup?"

  • "How do we handle payments?"

  • "Deployment process?"

Check Status:

  • "What do you know about this project?"

  • "Show me what context you have"

Update:

  • "Update: switched from Supabase to Firebase"

Critical Rules

1. Always Recall Before Answering

❌ Wrong: Makes assumptions about OAuth ✅ Right: "Recall authentication" → then answer with actual implementation

2. Never Assume Information

If no context stored, say so and ask for it. Record when provided.

3. Record Important Context Immediately

When you learn something about the codebase, record it right away.

4. Use Actual Context, Not Generic Examples

All examples should reflect your actual setup, not "ACME Corp" or generic code.

5. Check Status Regularly

"What do you know about this project?"

6. Update When Things Change

"Update: migrated to pgvector for vector storage"

7. Be Specific When Recording

"We use a database" "PostgreSQL 15 with pgvector, SQLAlchemy pool_size=20, AWS RDS with read replicas in us-east-1/us-west-2"

8. Cross-Reference Before Contradicting

If info conflicts with stored memories, recall and verify first.

9. Provide Feedback Regularly

"Please provide feedback: Memory Store correctly recalled our vector search. Rating: 9/10"

10. Use for Every Development Task

First stop for: new features, debugging, code review, tests, API integration, deployment, onboarding.

What to Record

Essential:

  • Business context and domain logic

  • Architecture decisions with reasoning

  • API contracts and integrations

  • Tech stack and infrastructure

  • Setup instructions and gotchas

  • Team conventions (commits, PRs, reviews)

  • CI/CD pipelines and deployment

  • Release processes and tagging

Bug Prevention:

  • Common mistakes and solutions

  • Past issues and fixes

  • Edge cases and gotchas

Cross-Repo:

  • Repository relationships

  • Shared types and packages

  • Integration points

Development Workflow Examples

Business Context

At [Company], we're building [Product]

Later: Help me implement the recall endpoint → AI recalls architecture, OAuth details, API patterns

Repository Context

Later: Add endpoint to search memories → AI generates code following your architecture

API Contracts

Later: Implement memory recording → AI uses correct endpoints, auth, error handling

Tech Stack

Local Setup

Cross-Repository

Use Cases Beyond Code

Infrastructure: Terraform patterns, Kubernetes configs, cloud services, Docker patterns

CI/CD: GitHub Actions/GitLab CI workflows, build processes, deployment automation, quality gates

Release Management: Staging → production workflows, tagging conventions, changelog generation, hotfix procedures

Organizational: Commit conventions, code review process, documentation standards, team workflows

Cross-Functional: Product requirements, design systems, security practices, monitoring standards

Best Practices

Do

  • Start with business context

  • Record architectural decisions with reasoning

  • Document API contracts immediately

  • Track setup gotchas

  • Build bug prevention database

  • Be explicit about critical info

  • Record cross-repo relationships

  • Update when things change

  • Document infrastructure/deployment

  • Store CI/CD configs

  • Track release conventions

  • Record team practices

  • Provide feedback regularly

Don't

  • Record sensitive data (API keys, passwords)

  • Dump entire codebases—focus on patterns

  • Record temporary details

  • Expect AI to infer what it wasn't told

How It Compounds

Week 1: Record initial context → AI starts understanding codebase

Month 1: Captures decisions and bug fixes → Prevents repeat mistakes

Month 3: Comprehensive knowledge base → Cross-repo understanding, team conventions documented

Month 6+: Institutional knowledge preserved → New members onboard instantly, AI becomes org expert

Time Savings: Infrastructure changes, CI/CD updates, release management, code reviews, onboarding, cross-project work, documentation—all faster with Memory Store.

Success Patterns

New Developer Onboarding: Record comprehensive guide once → New devs get instant, project-specific answers

Context Switching: AI maintains context across projects → Switch repos seamlessly, no context loss

Code Reviews: Record conventions → AI checks PRs against standards → Consistent quality

Bug Resolution: Auto-track bugs and solutions → AI recalls fixes → Team learning compounds

Infrastructure & DevOps: Record workflows → AI assists with changes, pipelines, releases → Context-aware help for Terraform, K8s, Docker

Release Management: Document process → AI helps create tags, generate notes, follow org process

Organizational Practices: Record conventions → AI ensures consistency across team and repos

Cross-Platform Memory

Memory Store works across: Cursor, Windsurf, Claude Code, Claude.ai, ChatGPT, Raycast AI, Poke, OpenCode, Codex, Droid, Goose, Gemini CLI, Qwen CLI

Record in Cursor → available in Claude.ai and vice versa.

Troubleshooting

Not Responding: Verify MCP extension installed → Check config → Restart Cursor → Ask "Is Memory Store working?"

Context Not Recalled: Ask explicitly "Recall [topic]" → Check what's stored → Record more specific context

Outdated: Record updates naturally: "Update: migrated from X to Y"

Performance: Use focused, specific context → Break long messages into chunks

Tips for Maximum Value

Start Projects Right: Spend 2-3 minutes recording what you're building, who it's for, tech stack, key constraints. Saves hours later.

Use as Project Notes: Instead of separate notes, use Memory Store as AI-accessible documentation.

Build Knowledge Base: Record insights and learnings as you go.

Use Across Workflow: Planning in Claude → Coding in Cursor → Troubleshooting in ChatGPT—all with same context.

Get Help

Discord: discord.gg/vynweB8qr3 - Support, workflows, community

Feedback: Share regularly via feedback tool:

Please provide feedback: [Your experience]

Tell us what works, what doesn't, feature requests, ratings.

This guide was created with Memory Store assistance.

‹ Memory Store for Raycast

Memory Store Prompt Templates ›

Everything everywhere, all at once.

Memory.Store