πŸ”’ Post-Quantum Encrypted Β· Local-First Β· Open Core

The Memory Layer
for AI Agents

Give your AI agents persistent, auditable, secure memory. UAML is the universal memory infrastructure that makes agents reliable.

Get Started View Source β†’
🧠

5 Memory Types

Episodic, semantic, procedural, reasoning, and associative memory β€” modeled after human cognition.

Details β†’
πŸ”

Post-Quantum Security

ML-KEM-768 encryption (NIST FIPS 203), key escrow with secret sharing, signed exports.

Details β†’
πŸ“‹

GDPR & ISO 27001 Ready

Built-in compliance auditor, consent tracking, DPIA generator, breach notification register.

Details β†’
🏠

Local-First

Your data stays on your hardware. SQLite-backed, zero cloud dependencies. Runs on Raspberry Pi.

Details β†’
πŸ”

Smart Recall

Policy-driven recall with context budgeting. Proactive memory surfacing β€” your agent anticipates what's relevant.

Details β†’
πŸ“Š

Full Audit Trail

Every read, write, and decision is logged. Reasoning traces link decisions to evidence. Provenance you can verify.

Details β†’
πŸ›‘οΈ

Security Configurator

Web UI wizard for firewall rules, AV exclusions, BitLocker, and WSL2 setup. One click to harden β€” no CLI knowledge needed.

Details β†’
πŸ€–

Expert on Demand

Temporarily grant your AI agent controlled access to the host OS. Time-limited, whitelisted, fully audited. Kill switch always available.

Details β†’

5-Layer Architecture

Data classified by sensitivity, sharing rules, and purpose β€” from personal identity to project deliverables. Learn more β†’

πŸͺͺ
Identity
Who am I?
Never shared
πŸ“š
Knowledge
What I know
Export only
πŸ‘₯
Team
Shared work
Access control
βš™οΈ
Operational
Logs & audit
Signed
πŸ“
Project
Deliverables
Centralized

πŸ›‘οΈ Security Configurator

The missing piece: a GUI tool for hardening the environment where your AI agent runs.

πŸ”₯

Firewall Setup

Configure UFW, iptables, Windows Firewall, or macOS pf β€” all from a visual wizard. No CLI required. The wizard generates OS-specific rules, shows them for review, and applies them with one click.

Details β†’
πŸ›‘οΈ

AV Exclusions

Windows Defender, Spotlight, ClamAV β€” exclude UAML data dirs from scanning. One click. Prevent false positives and improve performance without compromising security.

Details β†’
πŸ”

BitLocker VHD

Create an encrypted virtual disk for UAML data. XTS-AES-256. Recovery key management included. LUKS support on Linux. Your data stays protected even with physical disk access.

Details β†’
πŸ€–

Expert on Demand

Grant your AI temporary, audited access to the host OS. Time-limited sessions (max 60 min), command whitelist and blacklist, kill switch. Full audit trail. Risky actions require confirmation.

Details β†’
πŸ“„

Audit Reports

Every action logged with timestamp, command, and result. Export HTML reports for IT audit, management, or compliance documentation. Session history with filtering and search.

Details β†’
▢️

One-Click Execute

No copy-paste, no downloads. Click "Apply" β€” the command runs directly on your machine. Results shown instantly in a live log. The entire wizard runs locally on 127.0.0.1.

Details β†’
# Start the Security Configurator
from uaml.security.configurator import SecurityConfigurator

SecurityConfigurator().serve() # β†’ http://127.0.0.1:8785

# Open in browser β†’ wizard guides you through setup
# Click ▢️ Apply β†’ commands execute on your machine
# Download audit report β†’ send to your IT department

Learn more about Security Configurator β†’

Quick Start

Up and running in under a minute.

# Install
pip install uaml

# Initialize
uaml init

# Store knowledge
uaml learn "Python GIL was removed in 3.13 (free-threaded)"

# Search
uaml search "Python threading"

# Run compliance audit
uaml compliance audit

Python API

Full control from your code.

from uaml.core.store import MemoryStore

store = MemoryStore("memory.db")

# Learn
store.learn("UAML uses SQLite for local-first storage",
    topic="architecture", data_layer="knowledge")

# Recall with context budget
results = store.policy_recall("storage engine")

# Proactive memory
relevant = store.proactive_recall("discussing database options")

# Reasoning trace
store.capture_reasoning("Chose SQLite",
    reasoning="Zero dependencies, local-first")
<<<<<<< HEAD =======

Pricing

Open core. Start free, scale when you need to.

View our complete pricing with currency selection and annual billing options on our sales website.

View Pricing Plans β†’
>>>>>>> 923a61d77893b8840514c8e0579a200267548ce7