MemoryPilot

One memory. Every AI. Zero repetition.

MemoryPilot is a passive, intelligent memory layer for AI-assisted development. It automatically captures context from your work (git commits, file changes, terminal commands) and makes it available to any AI tool.

Your AI tools will finally remember you.

The Problem

You explain your project    → ChatGPT     → Session ends → FORGOTTEN
You explain it again        → Claude Code → Session ends → FORGOTTEN  
You explain it again        → Cursor      → Session ends → FORGOTTEN
You explain it again        → Gemini      → Session ends → FORGOTTEN

Same context. Explained 4 times. Every. Single. Day.

The Solution

MemoryPilot runs in the background, watching your development activity and building a persistent memory that follows you across all AI tools.

                    MemoryPilot
                   (One Memory)
                        │
        ┌───────┬───────┼───────┬───────┐
        ▼       ▼       ▼       ▼       ▼
    ChatGPT  Claude  Cursor  Gemini  Windsurf

Installation

# macOS/Linux
curl -fsSL https://contextpilot.dev/memorypilot/install.sh | sh

# Go install
go install github.com/memorypilot/memorypilot@latest

# From source
git clone https://github.com/memorypilot/memorypilot.git
cd memorypilot
go build -o memorypilot .

Quick Start

# Initialize MemoryPilot
memorypilot init

# Start the background daemon
memorypilot daemon start

# Check status
memorypilot status

# Search your memories
memorypilot recall "authentication patterns"

# Manually remember something
memorypilot remember --type decision "Chose PostgreSQL for ACID compliance"

MCP Integration (Claude Code, OpenClaw, Windsurf)

Add to your MCP configuration:

{
  "mcpServers": {
    "memorypilot": {
      "command": "memorypilot",
      "args": ["mcp"]
    }
  }
}

Then in your AI chat:

You: "How did we handle auth before?"
AI: [calls memorypilot_recall]
AI: "Based on your memory, you implemented OAuth2 with PKCE 
     for the mobile app on January 15th..."

Features

What MemoryPilot Captures

Source	What We Learn
Git commits	Decisions, patterns, history
File changes	Architecture evolution, refactors
Terminal commands	Workflows, tools, processes

Memory Types

Type	Description
decision	Architectural or technical choices
pattern	Recurring approaches or solutions
fact	Objective information
preference	Personal or team preferences
mistake	Errors to avoid
learning	New knowledge acquired

Privacy First

• Local-first: All data stored locally by default
• Smart filtering: Automatically redacts secrets and sensitive data
• No telemetry: Your memory is yours

Commands

memorypilot init          # Initialize MemoryPilot
memorypilot daemon start  # Start background daemon
memorypilot daemon stop   # Stop background daemon
memorypilot status        # Show status and statistics
memorypilot recall        # Search memories
memorypilot remember      # Manually create a memory
memorypilot mcp           # Start MCP server (for AI tool integration)

Configuration

Configuration file: ~/.memorypilot/config.yaml

# LLM for memory extraction
extraction:
  provider: ollama  # ollama | claude
  model: llama3.2

# Watchers
watchers:
  git:
    enabled: true
    interval: 30s
  file:
    enabled: true
    ignore: [node_modules, .git, dist]
  terminal:
    enabled: true

Roadmap

• Core agent with watchers
• SQLite memory store
• CLI commands
• MCP server
• Vector embeddings (semantic search)
• AI-powered memory extraction
• VS Code extension
• Web dashboard
• Cloud sync
• Team features

License

MIT

---

Built with ❤️ by Jitin Gambhir