MEMCORE — Universal AI Memory Layer

Git for AI memory. Give every AI tool persistent memory across sessions.

The problem

Every time you open Claude Code, Cursor, or ChatGPT — it forgets everything. Your architecture decisions. Your team conventions. Your project context. You explain the same things over and over. Every session. Every developer.

No AI tool has real persistent memory. Until now.

What is MEMCORE?

MEMCORE is a universal memory layer that works with any AI tool. Save context once. Every AI remembers it forever. ash npm install -g memcore

How it works

`ash

1. Initialize in your project

memcore init

2. Save what matters

memcore save

> Memory name: auth system

> Type: project

> Content: we use JWT, not sessions — compliance requirement

3. Inject into your AI tools

memcore inject # → writes CLAUDE.md for Claude Code memcore inject-cursor # → writes .cursorrules for Cursor memcore inject-chatgpt # → exports context for ChatGPT

4. Your AI now knows everything — without you explaining anything

claude # Claude Code reads CLAUDE.md automatically `

Team sharing

`ash

Dev 1 — save and share

memcore save memcore push git add memcore-sync.json && git commit -m "chore: update team memory"

Dev 2 — get all context instantly

git pull memcore pull memcore inject

Done. Full project context in seconds.

`

Commands

| Command | What it does | |---|---| | memcore init | Initialize in current project | | memcore save | Save a new memory | | memcore list | List all memories | | memcore search query | Search memories | | memcore delete id | Delete a memory | | memcore inject | Inject into CLAUDE.md | | memcore inject-cursor | Inject into .cursorrules | | memcore inject-chatgpt | Export for ChatGPT | | memcore push | Export team memories | | memcore pull | Import team memories | | memcore export | Export to memcore.json | | memcore import | Import from memcore.json |

Memory types

| Type | What to save | |---|---| | project | Architecture decisions, tech stack, constraints | | feedback | How to work with you, what to avoid | | user | Your role, expertise level, preferences | | reference | Where info lives — Linear, Jira, Slack, Grafana |

SDK

` ypescript import { memcore } from 'memcore'

// Get full project context as string const context = memcore.getContext()

// Save a memory programmatically memcore.save('auth system', 'project', 'we use JWT — compliance requirement')

// Search memories const results = memcore.search('auth')

// Inject into AI tools memcore.inject() // Claude Code memcore.injectCursor() // Cursor `

The memcore.json protocol

MEMCORE defines an open standard for AI memory. Any tool can read and write it. json { "version": "1.0", "project": { "name": "my-project" }, "memories": [ { "name": "auth system", "type": "project", "content": "we use JWT, not sessions" } ], "rules": ["never touch legacy/payments.ts"], "references": { "bugs": "Linear project INGEST" }, "integrations": { "claude": true, "cursor": true, "chatgpt": true } }

AI tools supported

| Tool | How | Status | |---|---|---| | Claude Code | Writes CLAUDE.md | Live | | Cursor | Writes .cursorrules | Live | | ChatGPT | Exports .txt | Live | | Gemini CLI | Coming soon | Planned | | GitHub Copilot | Coming soon | Planned |

Roadmap

• [x] Core memory system
• [x] Claude Code integration
• [x] Cursor integration
• [x] ChatGPT integration
• [x] Team sync via Git
• [x] Open protocol (memcore.json)
• [x] SDK
• [ ] Daemon (auto-injection)
• [ ] MEMCORE Cloud (team SaaS)
• [ ] Memory Graph
• [ ] Notion, Slack, Jira connectors

License

MIT — free forever for individuals.

Built with the belief that AI tools should remember what matters.