simplemindedbot/mnemex

Mnemex: Temporal Memory for AI

A Model Context Protocol (MCP) server providing human-like memory dynamics for AI assistants. Memories naturally fade over time unless reinforced through use, mimicking the Ebbinghaus forgetting curve.

[!WARNING] 🚧 ACTIVE DEVELOPMENT - EXPECT BUGS 🚧

This project is under active development and should be considered experimental. You will likely encounter bugs, breaking changes, and incomplete features. Use at your own risk. Please report issues on GitHub, but understand that this is research code, not production-ready software.

Known issues:

• API may change without notice between versions
• Test coverage is incomplete

📖 New to this project? Start with the for a simple explanation of what this does and how to use it.

Overview

This repository contains research, design, and a complete implementation of a short-term memory system that combines:

• Novel temporal decay algorithm based on cognitive science
• Reinforcement learning through usage patterns
• Two-layer architecture (STM + LTM) for working and permanent memory
• Smart prompting patterns for natural LLM integration
• Git-friendly storage with human-readable JSONL
• Knowledge graph with entities and relations

Why Mnemex?

🔒 Privacy & Transparency

All data stored locally on your machine - no cloud services, no tracking, no data sharing.

• Short-term memory: Human-readable JSONL files (~/.config/mnemex/jsonl/)

  • One JSON object per line
  • Easy to inspect, version control, and backup
  • Git-friendly format for tracking changes

• Long-term memory: Markdown files optimized for Obsidian

  • YAML frontmatter with metadata
  • Wikilinks for connections
  • Permanent storage you control

You own your data. You can read it, edit it, delete it, or version control it - all without any special tools.

Core Algorithm

The temporal decay scoring function:

$$ \Large \text{score}(t) = (n_{\text{use}})^\beta \cdot e^{-\lambda \cdot \Delta t} \cdot s $$

Where:

• $\large n_{\text{use}}$ - Use count (number of accesses)
• $\large \beta$ (beta) - Sub-linear use count weighting (default: 0.6)
• $\large \lambda = \frac{\ln(2)}{t_{1/2}}$ (lambda) - Decay constant; set via half-life (default: 3-day)
• $\large \Delta t$ - Time since last access (seconds)
• $\large s$ - Strength parameter $\in [0, 2]$ (importance multiplier)

Thresholds:

• $\large \tau_{\text{forget}}$ (default 0.05) — if score < this, forget
• $\large \tau_{\text{promote}}$ (default 0.65) — if score ≥ this, promote (or if $\large n_{\text{use}}\ge5$ in 14 days)

Decay Models:

• Power‑Law (default): heavier tail; most human‑like retention
• Exponential: lighter tail; forgets sooner
• Two‑Component: fast early forgetting + heavier tail

See detailed parameter reference, model selection, and worked examples in docs/scoring_algorithm.md.

Tuning Cheat Sheet

• Balanced (default)
  • Half-life: 3 days (λ ≈ 2.67e-6)
  • β = 0.6, τ_forget = 0.05, τ_promote = 0.65, use_count≥5 in 14d
  • Strength: 1.0 (bump to 1.3–2.0 for critical)
• High‑velocity context (ephemeral notes, rapid switching)
  • Half-life: 12–24 hours (λ ≈ 1.60e-5 to 8.02e-6)
  • β = 0.8–0.9, τ_forget = 0.10–0.15, τ_promote = 0.70–0.75
• Long retention (research/archival)
  • Half-life: 7–14 days (λ ≈ 1.15e-6 to 5.73e-7)
  • β = 0.3–0.5, τ_forget = 0.02–0.05, τ_promote = 0.50–0.60
• Preference/decision heavy assistants
  • Half-life: 3–7 days; β = 0.6–0.8
  • Strength defaults: 1.3–1.5 for preferences; 1.8–2.0 for decisions
• Aggressive space control
  • Raise τ_forget to 0.08–0.12 and/or shorten half-life; schedule weekly GC
• Environment template
  • MNEMEX_DECAY_LAMBDA=2.673e-6, MNEMEX_DECAY_BETA=0.6
  • MNEMEX_FORGET_THRESHOLD=0.05, MNEMEX_PROMOTE_THRESHOLD=0.65
  • MNEMEX_PROMOTE_USE_COUNT=5, MNEMEX_PROMOTE_TIME_WINDOW=14

Decision thresholds:

• Forget: $\text{score} < 0.05$ → delete memory
• Promote: $\text{score} \geq 0.65$ OR $n_{\text{use}} \geq 5$ within 14 days → move to LTM

Key Innovations

1. Temporal Decay with Reinforcement

Unlike traditional caching (TTL, LRU), memories are scored continuously based on:

• Recency - Exponential decay over time
• Frequency - Use count with sub-linear weighting
• Importance - Adjustable strength parameter

This creates memory dynamics that closely mimic human cognition.

2. Smart Prompting System

Patterns for making AI assistants use memory naturally:

Auto-Save

User: "I prefer TypeScript over JavaScript"
→ Automatically saved with tags: [preferences, typescript, programming]

Auto-Recall

User: "Can you help with another TypeScript project?"
→ Automatically retrieves preferences and conventions

Auto-Reinforce

User: "Yes, still using TypeScript"
→ Memory strength increased, decay slowed

No explicit memory commands needed - just natural conversation.

3. Two-Layer Architecture

┌─────────────────────────────────────┐
│   Short-term memory                 │
│   - JSONL storage                   │
│   - Temporal decay                  │
│   - Hours to weeks retention        │
└──────────────┬──────────────────────┘
               │ Automatic promotion
               ↓
┌─────────────────────────────────────┐
│   LTM (Long-Term Memory)            │
│   - Markdown files (Obsidian)       │
│   - Permanent storage               │
│   - Git version control             │
└─────────────────────────────────────┘

Project Structure

mnemex/
├── README.md                          # This file
├── CLAUDE.md                          # Guide for AI assistants
├── src/mnemex/
│   ├── core/                          # Decay, scoring, clustering
│   ├── storage/                       # JSONL and LTM index
│   ├── tools/                         # 10 MCP tools
│   ├── backup/                        # Git integration
│   └── vault/                         # Obsidian integration
├── docs/
│   ├── scoring_algorithm.md           # Mathematical details
│   ├── prompts/                       # Smart prompting patterns
│   ├── architecture.md                # System design
│   └── api.md                         # Tool reference
├── tests/                             # Test suite
├── examples/                          # Usage examples
└── pyproject.toml                     # Project configuration

Quick Start

Installation

Recommended: UV Tool Install

# Install from GitHub (recommended)
uv tool install git+https://github.com/simplemindedbot/mnemex.git

# Or install from local directory (for development)
uv tool install .

This installs mnemex and all 7 CLI commands as isolated tools.

Alternative: Editable Install (for development)

# Clone and install in editable mode
git clone https://github.com/simplemindedbot/mnemex.git
cd mnemex
uv pip install -e ".[dev]"

Configuration

Copy .env.example to .env and configure:

# Storage
MNEMEX_STORAGE_PATH=~/.config/mnemex/jsonl

# Decay model (power_law | exponential | two_component)
MNEMEX_DECAY_MODEL=power_law

# Power-law parameters (default model)
MNEMEX_PL_ALPHA=1.1
MNEMEX_PL_HALFLIFE_DAYS=3.0

# Exponential (if selected)
# MNEMEX_DECAY_LAMBDA=2.673e-6  # 3-day half-life

# Two-component (if selected)
# MNEMEX_TC_LAMBDA_FAST=1.603e-5  # ~12h
# MNEMEX_TC_LAMBDA_SLOW=1.147e-6  # ~7d
# MNEMEX_TC_WEIGHT_FAST=0.7

# Common parameters
MNEMEX_DECAY_LAMBDA=2.673e-6
MNEMEX_DECAY_BETA=0.6

# Thresholds
MNEMEX_FORGET_THRESHOLD=0.05
MNEMEX_PROMOTE_THRESHOLD=0.65

# Long-term memory (optional)
LTM_VAULT_PATH=~/Documents/Obsidian/Vault

MCP Configuration

Add to your Claude Desktop config (~/Library/Application Support/Claude/claude_desktop_config.json):

{
  "mcpServers": {
    "mnemex": {
      "command": "mnemex"
    }
  }
}

That's it! No paths, no environment variables needed.

For development (editable install):

{
  "mcpServers": {
    "mnemex": {
      "command": "uv",
      "args": ["--directory", "/path/to/mnemex", "run", "mnemex"],
      "env": {"PYTHONPATH": "/path/to/mnemex/src"}
    }
  }
}

Configuration:

• Storage paths are configured in ~/.config/mnemex/.env or project .env
• See .env.example for all available settings

Maintenance

Use the maintenance CLI to inspect and compact JSONL storage:

# Show storage stats (active counts, file sizes, compaction hints)
mnemex-maintenance stats

# Compact JSONL (rewrite without tombstones/duplicates)
mnemex-maintenance compact

Migrating to UV Tool Install

If you're currently using an editable install (uv pip install -e .), you can switch to the simpler UV tool install:

# 1. Uninstall editable version
uv pip uninstall mnemex

# 2. Install as UV tool
uv tool install git+https://github.com/simplemindedbot/mnemex.git

# 3. Update Claude Desktop config to just:
#    {"command": "mnemex"}
#    Remove the --directory, run, and PYTHONPATH settings

Your data is safe! This only changes how the command is installed. Your memories in ~/.config/mnemex/ are untouched.

Migrating from STM Server

If you previously used this project as "STM Server", use the migration tool:

# Preview what will be migrated
mnemex-migrate --dry-run

# Migrate data files from ~/.stm/ to ~/.config/mnemex/
mnemex-migrate --data-only

# Also migrate .env file (rename STM_* variables to MNEMEX_*)
mnemex-migrate --migrate-env --env-path ./.env

The migration tool will:

• Copy JSONL files from ~/.stm/jsonl/ to ~/.config/mnemex/jsonl/
• Optionally rename environment variables (STM_* → MNEMEX_*)
• Create backups before making changes
• Provide clear next-step instructions

After migration, update your Claude Desktop config to use mnemex instead of stm.

CLI Commands

The server includes 7 command-line tools:

mnemex                  # Run MCP server
mnemex-migrate          # Migrate from old STM setup
mnemex-index-ltm        # Index Obsidian vault
mnemex-backup           # Git backup operations
mnemex-vault            # Vault markdown operations
mnemex-search           # Unified STM+LTM search
mnemex-maintenance      # JSONL storage stats and compaction

MCP Tools

10 tools for AI assistants to manage memories:

Example: Unified Search

Search across STM and LTM with the CLI:

mnemex-search "typescript preferences" --tags preferences --limit 5 --verbose

Example: Reinforce (Touch) Memory

Boost a memory's recency/use count to slow decay:

{
  "memory_id": "mem-123",
  "boost_strength": true
}

Sample response:

{
  "success": true,
  "memory_id": "mem-123",
  "old_score": 0.41,
  "new_score": 0.78,
  "use_count": 5,
  "strength": 1.1
}

Example: Promote Memory

Suggest and promote high-value memories to the Obsidian vault.

Auto-detect (dry run):

{
  "auto_detect": true,
  "dry_run": true
}

Promote a specific memory:

{
  "memory_id": "mem-123",
  "dry_run": false,
  "target": "obsidian"
}

As an MCP tool (request body):

{
  "query": "typescript preferences",
  "tags": ["preferences"],
  "limit": 5,
  "verbose": true
}

Example: Consolidate Similar Memories

Find and merge duplicate or highly similar memories to reduce clutter:

Auto-detect candidates (preview):

{
  "auto_detect": true,
  "mode": "preview",
  "cohesion_threshold": 0.75
}

Apply consolidation to detected clusters:

{
  "auto_detect": true,
  "mode": "apply",
  "cohesion_threshold": 0.80
}

The tool will:

• Merge content intelligently (preserving unique information)
• Combine tags and entities (union)
• Calculate strength based on cluster cohesion
• Preserve earliest created_at and latest last_used timestamps
• Create tracking relations showing consolidation history

Mathematical Details

Decay Curves

For a memory with $n_{\text{use}}=1$, $s=1.0$, and $\lambda = 2.673 \times 10^{-6}$ (3-day half-life):

Use Count Impact

With $\beta = 0.6$ (sub-linear weighting):

Frequent access significantly extends retention.

Documentation

• - Complete mathematical model with LaTeX formulas
• - Patterns for natural LLM integration
• - System design and implementation
• - MCP tool documentation
• - Knowledge graph usage

Use Cases

Personal Assistant (Balanced)

• 3-day half-life
• Remember preferences and decisions
• Auto-promote frequently referenced information

Development Environment (Aggressive)

• 1-day half-life
• Fast context switching
• Aggressive forgetting of old context

Research / Archival (Conservative)

• 14-day half-life
• Long retention
• Comprehensive knowledge preservation

License

MIT License - See for details.

Clean-room implementation. No AGPL dependencies.

Related Work

• Model Context Protocol - MCP specification
• Ebbinghaus Forgetting Curve - Cognitive science foundation
• Research inspired by: Memoripy, Titan MCP, MemoryBank

Citation

If you use this work in research, please cite:

@software{mnemex_2025,
  title = {Mnemex: Temporal Memory for AI},
  author = {simplemindedbot},
  year = {2025},
  url = {https://github.com/simplemindedbot/mnemex},
  version = {1.0.0}
}

Contributing

Contributions are welcome! See for detailed instructions.

🚨 Help Needed: Windows & Linux Testers!

I develop on macOS and need help testing on Windows and Linux. If you have access to these platforms, please:

• Try the installation instructions
• Run the test suite
• Report what works and what doesn't

See the in CONTRIBUTING.md for details.

General Contributions

For all contributors, see for:

• Platform-specific setup (Windows, Linux, macOS)
• Development workflow
• Testing guidelines
• Code style requirements
• Pull request process

Quick start:

1. Read for platform-specific setup
2. Understand the
3. Review the
4. Follow existing code patterns
5. Add tests for new features
6. Update documentation

Status

Version: 1.0.0 Status: Research implementation - functional but evolving

Phase 1 (Complete) ✅

• 10 MCP tools

• Temporal decay algorithm

• Knowledge graph

Phase 2 (Complete) ✅

• JSONL storage
• LTM index
• Git integration
• Smart prompting documentation
• Maintenance CLI
• Memory consolidation (algorithmic merging)

Future Work

• Spaced repetition optimization
• Adaptive decay parameters
• Performance benchmarks
• LLM-assisted consolidation (optional enhancement)

Built with Claude Code 🤖