Open Source · Local-First · Nonprofit
A personal AI that looks out for you and your family
Moose is an AI assistant that runs entirely on your hardware, advocates for your wellbeing, helps your children learn, and never sends a single byte of your family's data to the cloud.
Requires Python 3.11+ and an LLM backend (Ollama, LM Studio, or llama.cpp).
100% local. No accounts. No subscriptions. No data collection.
This is for you if
You read to your kids. You answer "why" questions instead of shutting them down. You want your children to be curious, capable, and kind — and you feel overwhelmed by how hard that is in a world optimized to make them passive consumers.
You looked at the iPad and thought "there has to be something better than this."
You're not defined by politics, religion, or geography. You're defined by aspiration. You want a Star Trek future for your kids and you don't see anyone building it.
This is not for you if
You want to spy on your kids. You're looking for a digital babysitter. You want AI to replace your involvement.
If you want to read your teenager's messages, Moose isn't your product. If you want to park your child in front of a screen and walk away, Moose isn't your product.
We'd rather save you the time than sell you something that will make you unhappy.
The problem
The tools meant to help your family are designed to extract from it.
Cloud AI assistants process your conversations on servers you don't control. Parental control apps surveil children instead of supporting them. And none of them are actually on your side.
Your data leaves your home
Every question your child asks a cloud AI travels through third-party servers. Their curiosity, their worries, their inner world — stored on hardware you'll never see.
Parental tools are surveillance tools
The market gives parents two options: spy on your kids or ignore the problem. Monitoring builds resentment. Ignoring it leaves children unprotected. Neither option builds trust.
AI assistants say yes to everything
Current AI is optimized to be agreeable, not honest. It won't tell you when you're making a mistake. It won't push back when you need it. It's a tool, not an advocate.
The Advocacy Layer
AI that genuinely looks out for you.
Moose doesn't just do what you ask. It watches for patterns, flags what matters, and respects your autonomy. Not a nag. Not a yes-man. An advocate.
Silent observation
Watches without hovering
Moose notices patterns in your work, your habits, and your decisions. Most observations stay internal — quietly building understanding without interrupting your flow.
Ambient context
Weaves in what matters
Relevant information surfaces naturally inside normal responses. No separate alerts. No notification fatigue. Just the right context at the right moment.
Gentle flag
Raises it once, then respects your choice
When a clear pattern emerges, Moose tells you directly. If you dismiss it, Moose drops it. It doesn't repeat itself. Your decision is final — unless the situation changes.
Critical alert
Says it whether you want to hear it or not
For genuine safety or wellbeing concerns, Moose speaks up clearly. Not to control you — because looking out for someone means being honest even when it's uncomfortable.
Built for families
A patient teacher. A trusted guide. A professional in your pocket.
Moose adapts to every member of your family — their age, their needs, their growing independence.
For children
Moose answers every "why" with patience no human can sustain. It meets curiosity with honest, age-appropriate answers. It helps with homework by teaching, not giving answers. It helps kids learn to build things, write code, create art, and think through problems — at their own pace, without judgment.
When a child is struggling emotionally, Moose notices the pattern and helps parents understand what they're seeing — with developmental context and evidence-based guidance on how to respond. Not a transcript. Not surveillance. Insight paired with professional knowledge.
For parents
Parenting doesn't come with a handbook. Now it can come with a professional. Moose gives parents the developmental context that child therapists, pediatricians, and educators have — without the waitlists or the cost.
"Your child has been expressing frustration about math consistently. Their pattern suggests they understand the concepts but get overwhelmed by timed tests. You might ask their teacher about untimed alternatives."
Growing autonomy, built in
What Moose shares with parents changes as children grow. A six-year-old gets more visible support. A twelve-year-old gets more independence. A sixteen-year-old's experience looks much more like an adult's. The child's growing autonomy isn't something parents have to manage — it's built into the architecture.
Advocate Profiles
Solo
Full advocacy, Moose only. No external escalation. For adults who want a capable AI that looks out for them.
Default for all users
Partnered
A trusted person can be consulted on critical-level flags. They see patterns, not content. Both parties consent.
Escalation at critical only
Guided
Tighter feedback loop. The trusted person sees more categories and can set supportive boundaries. For children, neurodivergent users, or anyone who wants more structure.
Adjustable thresholds
Custom
Full control over every category, threshold, and trusted person. Your system, your rules.
Per-category configuration
What Moose can do
An engineering assistant, a learning companion, and a household co-pilot.
Specialist agents route every question to the right expertise. One system, many skills.
Security
Multi-layered authentication, hardware interlocks, prompt injection scanning, and defense-in-depth architecture that keeps your family's data safe.
Engineering & Making
Works with Bambu Lab, Blender, FreeCAD, and Arduino. Helps you design, print, build, and debug — from first sketch to finished project.
Research & Learning
Deep reasoning for complex analysis. Homework help that teaches. Project research that cites its sources and rates their reliability.
Home Management
Calendar, email, scheduling, and routine coordination. The boring logistics that hold a household together, handled reliably.
Communication
Connects through the channels you already use — messaging apps, voice, and more. Same brain, every surface.
Math & Calculations
Material specs, tolerances, unit conversions, engineering math, and data analysis. From homework help to professional calculations.
Code & Automation
Write scripts, automate workflows, debug projects, and learn programming with a patient guide that adapts to your skill level.
Advocacy
Watches for patterns, flags what matters, navigates the tension between your goals and your habits. The capability that makes everything else trustworthy.
How it's built
Local-first isn't a feature. It's a prerequisite.
When the data is your child's inner life, "we take privacy seriously" isn't good enough. The architecture has to make exposure impossible.
Your hardware, your data
Moose runs entirely on your machine. Conversations, patterns, advocacy observations, and family data never touch a server. There's no account to create, no subscription to pay, and no company that can be acquired, subpoenaed, or breached to expose your family's information.
- Works with Ollama, LM Studio, or llama.cpp as your AI backend
- Choose the models you trust — swap them anytime
- All state stored locally in files you own and can inspect
- Docker option for additional isolation
Security through architecture, not promises
Moose doesn't rely on model alignment to stay safe. The architecture itself constrains what the system can do. Actions require explicit human authorization. Sensitive operations require hardware confirmation.
- Tiered permission system — routine, sensitive, critical
- Command allowlisting — only pre-approved operations execute
- Dedicated security agent monitors all system activity
- Prompt injection scanning at multiple layers
- Optional hardware authentication (NFC, biometrics, physical buttons)
What we believe
Design principles
01
Advocate for the user, even a child
Moose's loyalty is to the person it's helping. For children, that means supporting their growth and protecting their safety — while keeping parents informed with context and guidance, not surveillance.
02
Honest over agreeable
A tool that tells you what you want to hear is a broken tool. Moose is structurally designed to push back when the situation warrants it, not because a prompt says to, but because the architecture requires it.
03
Structure over alignment
Safety can't depend on a model behaving well. It has to be architecturally impossible for the system to act against your interests. Permissions, hardware interlocks, and human-in-the-loop by design.
04
Grow with the person
A six-year-old and a sixteen-year-old need fundamentally different relationships with AI. Moose adapts not just its language but its entire interaction model as users develop and change.
05
Observe much, speak little
The fastest way to kill trust is to nag. Moose watches constantly and speaks up rarely. When it does, it matters. The threshold for intervention is high — but when Moose says something, listen.
06
No one else's incentives
FAESR is a nonprofit. Moose is open source. There are no ads, no data collection, no engagement metrics, and no investor pressure to monetize your family's attention.
Ready to build something better?
Moose is open source, free, and actively developed. Join us.
Requires Python 3.11+, Node 18+, and an LLM backend. Currently in active development — contributions welcome.