@memor-dev/memor

Local memory persistence for AI coding assistants.

Every AI coding tool — Copilot, Claude Code, Cursor, Windsurf — starts every conversation cold. Memor fixes that. It stores project context locally, indexes it with a full search engine, and surfaces only the relevant memories within a token budget.

Five text files per project. Full indexing engine. Zero cloud, zero daemon, zero git commits.

Install

npm i -g @memor-dev/memor

Or run directly:

npx @memor-dev/memor init

Quick Start

# Initialize in your project
cd your-project
memor init

memor init creates .memor/, injects copilot-instructions.md and .github/skills/memor/SKILL.md so your AI tool automatically reads and writes memories. No extra setup needed.

Use memor init --tools claude,cursor,windsurf to configure additional AI tools.

Additional Commands

# Add a memory manually
memor add -s "#arch #db: PostgreSQL 16 with Drizzle ORM"

# Get context for AI tools
memor context --budget 10000

# Search memories
memor search "deploy"

# Compact WAL into snapshot
memor compact

How It Works

Conversations ──► APPEND to memory.wal (JSONL)
                       │
                       ▼
                  COMPACTION (score → dedupe → budget)
                       │
                       ▼
                  memory.db (compact DSL, token-budgeted)
                       │
                       ▼
             AI tools READ memory.db at conversation start

• Write path: Append memories as JSONL — fast, no coordination
• Read path: Trigram index + BM25 ranking for sub-millisecond retrieval
• Compaction: Deduplicates, scores, enforces token budget

Supported Platforms

| OS | Architecture | |---|---| | Linux | x64, arm64 | | macOS | x64, arm64 (Apple Silicon) | | Windows | x64 |

Documentation

Full documentation, design details, and CLI reference at github.com/akashchekka/memor.

License

MIT