Mnemonic

Mnemonic — Unified Memory System

Mnemonic aggregates your communications and AI sessions across all channels (iMessage, Gmail, Cursor, Codex, etc.) into a unified searchable memory with identity resolution, semantic search, and analysis.

Why Mnemonic?

Your communications and AI interactions are fragmented:

• iMessage threads with some people
• Email conversations with others
• AI chat sessions in Cursor
• More AI sessions in Codex

Mnemonic unifies them into one memory layer, so you can:

• Search across ALL channels semantically
• Extract memories and entities automatically
• Enable smart forking for AI sessions
• Query your complete communication history

Architecture

Mnemonic uses a ledger-based architecture:

• Core Ledger — Shared infrastructure: episodes, analysis runs, facets, embeddings
• Events Ledger — Human communications: iMessage, Gmail, calendar, trimmed AI turns
• Agents Ledger — Full fidelity AI sessions: messages, turns, tool calls (for smart forking)

Quick Start

# Initialize
mnemonic init

# Configure your identity
mnemonic me set --name "Your Name" --phone "+1234567890" --email "you@example.com"

# Connect adapters
mnemonic connect imessage
mnemonic connect gmail --account you@gmail.com
mnemonic connect cursor  # AI sessions via AIX

# Sync all channels
mnemonic sync

# Query
mnemonic events --person "Dad" --since "2025-01-01"
mnemonic search "authentication flow"
mnemonic people --top 20

Commands

Setup

Sync

Query

Identity Management

Adapters

iMessage (via Eve)

Prerequisites:

brew install Napageneral/tap/eve
eve init && eve sync

Connect:

mnemonic connect imessage

Gmail (via gogcli)

Prerequisites:

brew install steipete/tap/gogcli
gog auth add your@gmail.com

Connect:

mnemonic connect gmail --account your@gmail.com

AI Sessions (via AIX)

Connect:

mnemonic connect cursor

This imports AI sessions from Cursor via AIX:

• Events Ledger: Trimmed turns (user message + response)
• Agents Ledger: Full fidelity (all messages, tool calls, turns)

Output Formats

All commands support --json / -j:

mnemonic events --json | jq '.events[] | select(.channel == "imessage")'
mnemonic people --top 10 --json

Configuration

Config: ~/.config/mnemonic/config.yaml

me:
  canonical_name: "Your Name"
  identities:
    - channel: imessage
      identifier: "+1234567890"
    - channel: email
      identifier: "you@example.com"

adapters:
  imessage:
    type: eve
    enabled: true
  gmail:
    type: gogcli
    enabled: true
    account: you@gmail.com
  cursor:
    type: aix
    enabled: true

Data: ~/Library/Application Support/Mnemonic/mnemonic.db

Bootstrap (for AI agents)

# Check if installed
which mnemonic && mnemonic version

# Install
brew install Napageneral/tap/mnemonic
# OR: go install github.com/Napageneral/mnemonic/cmd/mnemonic@latest

# Setup
mnemonic init

# Configure identity
mnemonic me set --name "User Name" --email "user@example.com"

# Connect adapters (assumes Eve/gogcli/AIX already set up)
mnemonic connect imessage
mnemonic connect gmail --account user@gmail.com
mnemonic connect cursor

# Sync
mnemonic sync

# Verify
mnemonic db query "SELECT COUNT(*) as count FROM events"
mnemonic people --top 5

Tips for Agents

1. Use mnemonic people --top 10 to understand who the user communicates with most
2. Use mnemonic events --person "Name" to get context on a relationship
3. Use mnemonic search "topic" for semantic search across all content
4. Use mnemonic timeline --today for recent activity
5. Filter by channel to focus on specific contexts
6. Use --json output for programmatic access
7. Raw SQL via mnemonic db query for complex queries