AegisKids tablet OEM
Lock down on-device assistants out of the box.
Route every prompt through Aegis at the OS level. Ship a kid-safe assistant on day one without retraining the underlying model.
POST /v1/aegis/filter Meet Aegis 2.0
The Model that makes all AI models kid-safe.
Aegis 2.0 sits in front of any model - OpenAI, Anthropic, Gemini, Llama, or your own fine-tune - filtering and rewriting outputs in real time so they're age-appropriate.
Llama 4Mistral LargeDeepSeek V3Qwen 3Phi-4Command R+GLM 5GPT-5Claude 4Gemini 2.5Grok 3Kimi K2Reka CoreJamba 1.5
Aegis 2.0
Llama 4Mistral LargeDeepSeek V3Qwen 3Phi-4Command R+GLM 5
Aegis 2.0
GPT-5Claude 4Gemini 2.5Grok 3Kimi K2Reka CoreJamba 1.5
Every prompt and every model response passes through Aegis. Unsafe content is rewritten or blocked before it ever reaches your user, with a full audit trail.
What Aegis does
Six jobs, one model, every query.
Four years of refinement, now public. Aegis ships the fundamentals out of the box and the tooling for parents and KidTech builders to stay in control of how AI behaves.
Safety
Hard guardrails on the topics that matter.
Some topics should never be broached. Aegis detects, short-circuits and circumvents these before a probabilistic model ever gets a chance to fumble the nuance - deterministic blocks on the highest-risk categories, not left to model whim. Yes, it's sometimes boring. That's the point: dangerous outputs aren't a coin flip.
One drop-in SDK. Every job above runs automatically on every call - no extra integration work.
Everything you need to ship kid-safe AI.
Six years of building consumer kid products, distilled into one drop-in SDK.
Age-aware moderation
Tone, topic, and reading-level checks tuned to 5-7, 8-11, and 12-15 brackets - not a generic adult filter.
Layered safety system
Aegis combines deterministic blocks, age-aware checks, and policy-tuned rewrites so risky content is caught before it reaches a child.
COPPA & GDPR-K compliant
Audit logs, data residency, and parent consent flows built in. Ship to regulated markets on day one.
Model-agnostic
OpenAI, Anthropic, Gemini, Llama, Mistral, your own fine-tune. One SDK, any provider.
Real-time rewriting
Soft-rewrite unsafe outputs into kid-friendly language instead of blunt refusals.
Built-in escalation
Parents and moderators are alerted in real time on potentially dangerous interactions, with the context they need to intervene fast.
Developer-ready
Drop Aegis in front of any model in an afternoon.
One REST endpoint, one SDK install. All six jobs run automatically on every call - no retraining, no model swap, no glue code. Self-host or use our managed cloud. Works with OpenAI, Anthropic, Gemini, Llama or your own fine-tune.
aegis-quickstart.ts
// 1. Install
$ npm install @astrosafe/aegis
// 2. Wrap any model call
import { Aegis } from "@astrosafe/aegis";
const aegis = new Aegis({
apiKey: "sk_***",
ageBand: "8-11",
model: "gpt-5",
});
const reply = await aegis.chat({
prompt: "What's a black hole?",
});
// → safety + triage + age-tuned reply
// + audit log, in one round-trip.Avg 92ms · 6 jobs · 1 call
More than a model - a full kid-safe platform.
Teams that adopt Aegis get the whole supporting cast - apps, dashboards, reporting, and a launch team to back them up.
White-label parent portal
A production-ready parent dashboard - activity, time controls, approvals and transparency. Drop it in under your brand, your colours, your domain.
White-label kid-facing apps
Pre-built kid apps for search, chat, video and learning - fully re-skinnable. Launch a full kid-safe product without rebuilding the front end.
Internal moderation dashboards
White-label tools for your trust & safety team - queues, escalations, age-bracket policy controls, and live model overrides.
Reporting & analytics
Out-of-the-box dashboards for safety incidents, intervention rates, engagement and parent-visible metrics - exportable to your BI stack.
Auditable documentation
We generate the policy, DPIA and model-card paperwork as you implement - so COPPA, GDPR-K and EU AI Act audits become a checkbox, not a project.
Embedded launch team
Our solutions engineers and child-safety experts sit alongside your team through integration, soft launch and scale - not just a Slack channel.
Use cases
What teams build with Aegis.
One API. Drop it in front of any model. Here's how kid-tech teams wire Aegis into the products they ship.
AegisEdTech app
A homework helper that won't go off the rails.
Age-aware tone, no exam cheating, no off-topic detours. Rewrite vague answers into curriculum-appropriate explanations.
aegis.rewrite({ ageBand: '8-12' })
AegisFamily streaming
Moderate AI-generated recommendations & summaries.
Score titles and synopses against your household age profile before they ever appear on the home rail.
aegis.score(content, profile)
AegisSmart toys
Conversational plushies that stay in character.
Hard guardrails on sensitive topics, persona drift detection, and full transcripts shipped to the parent app.
aegis.guard(turn, persona)
AegisParent companion
Give parents the receipts - and the controls.
Stream Aegis events into your parent dashboard. Show what was blocked, why, and let parents tune the dial per child.
GET /v1/aegis/audit
AegisSchools & districts
COPPA, GDPR-K and EU AI Act, audit-ready.
SSO with your district IdP, classroom-level policies, and exportable audit logs for every AI interaction.
aegis.policy.apply(school) Why AstroSafe
Generic AI vendors don't understand kids. We do!
Built on years of shipping safe consumer products for families. Tell us what you're building and we'll send a free AI safety read-out within 48 hours.