Atlas K-12

MTSS for the schools that matter most

A complete student-success platform with a local AI assistant trained by your teachers. Four whole-child pillars. 788 evidence-based techniques. Zero outbound network calls. Built to run on hardware your district already owns.

0
Curated Techniques
0
Whole-Child Pillars
0
Compass AI Tiers
0
Outbound Calls
Atlas K-12

Atlas K-12

Built On Your Server. Trained By Your Team.

Academics
|
Attendance
|
Behavior
|
Wellness
Under Development

Introducing Atlas K-12

Atlas K-12 is a proactive, data-informed student support platform built around the Multi-Tiered System of Supports (MTSS) framework. Designed for educators, counselors, administrators, and student support teams, Atlas K-12 brings your school's academic, behavioral, attendance, and social-emotional data together in one place so that every adult in the building can see the whole child, act on what the data shows, and ensure no student waits for support until it is too late.

Atlas K-12 is not a replacement for the systems your school already uses. It is the connective layer between them: a unified platform that speaks the language of MTSS, surfaces the right student information at the right time, and empowers your Student Support Teams to make data-based decisions with confidence.

MTSS is the gold standard for how schools address intervention and support, and Atlas K-12 is built to make that framework work for every student in your building, from Tier 1 universal instruction all the way through Tier 3 intensive, individualized support.

Every student. Every signal. One Care Team.

Atlas K-12 reads every academic, attendance, behavior, and social-emotional signal across your district, builds an intervention plan the moment a student needs one, surfaces it to a coordinated Care Team, tracks progress on the same screen, and updates every adult in the building in real time. No student waits. No teacher guesses. No system works alone.

Your Systems. Our Platform.

Schools do not lack data. They lack a way to bring it all together. Atlas K-12 connects to the platforms your district already uses, no migrations, no retraining, no disruption.

Student Information Systems

PowerSchool Infinite Campus Skyward (Qmlativ / SMS) Aeries Synergy (Edupoint) FACTS SIS Blackbaud K-12 Tyler SIS Genesis Focus School Software SchoolTool TxEIS / Ascender eSchoolPlus Aspen JCampus Frontline Illuminate Education

Learning Management Systems

Canvas Blackboard Moodle Google Classroom Schoology Jumprope

Assessment and SEL Screeners

NWEA MAP i-Ready Renaissance STAR Acadience FastBridge Panorama BIMAS DESSA PBIS Rewards State Assessments (all 50 states)

Data Standards and Import Methods

Ed-Fi Data Standard OneRoster (1EdTech) CSV / Excel Upload SFTP Nightly Sync REST API (OAuth 2.0) Open API Direct SQL Connection Flexible Field Mapping SSO (Google / Microsoft / Clever)

Atlas K-12: Where your data finally works together.

The Four Whole-Child Data Pillars

MTSS is a whole-child framework. Every pillar is equally valid and equally important.

📚

Academic Performance

Universal screener scores, formative assessment trends, grade-level benchmark alignment, and progress monitoring trajectories. Supports traditional letter grades and standards-based grading.

📅

Attendance and Engagement

Daily attendance tracking, chronic absenteeism identification at the 10 percent threshold, excused vs unexcused breakdown, and tardiness patterns.

📋

Behavioral Support Indicators

Office discipline referral patterns, PBIS data, CICO participation. Surfaces support needs through a strengths-based lens, never used to label students.

💖

Social-Emotional Wellbeing

Research-backed SEL screeners measuring grit, growth mindset, self-management, social awareness, self-efficacy, and emotion regulation.

Early Warning System

Proactive support identification across all four pillars. Atlas K-12 continuously monitors academic performance, attendance, behavior, and social-emotional data to surface students who may benefit from additional support before gaps widen. Every student is assigned a tier status: Universal, Targeted, or Intensive, aligned directly to the MTSS framework. No waiting. No guessing. The right support signal, at the right time.

TIER 3 | INTENSIVE ~5% of students TIER 2 | TARGETED ~15% of students TIER 1 | UNIVERSAL ~80% of students
TIER 1: UNIVERSAL (~80%)

All students receive high-quality, evidence-based core instruction and universal behavioral support. Includes school-wide PBIS frameworks. Universal screening conducted at least three times per year.

TIER 2: TARGETED (~15%)

Targeted, supplemental support delivered in small groups for students not yet demonstrating expected progress in Tier 1. Continuous progress monitoring over six to eight week intervention rounds.

TIER 3: INTENSIVE (~5%)

Intensive, individualized intervention for students who have not demonstrated expected progress through Tier 2 supports. Coordinated by a Care Team with high-frequency progress monitoring.

Tier distribution based on the OSEP Center on PBIS and National Center on Intensive Intervention 80/15/5 model.

Student Data Dashboard

Every data point that matters for a student, in one place. Academic performance, attendance, behavior, social-emotional indicators, intervention plans, referrals, Care Team notes, and demographics. Connected directly to your existing systems.

MTSS Intervention Planning

Create Tier 2 and Tier 3 intervention plans with goals, timelines, strategies, and progress monitoring. Goal graphs show if students are on pace. Champion assignment and documentation built in.

SEL Survey System

Research-backed social-emotional learning surveys for grades 3-5 and 6-12. Measures grit, growth mindset, self-management, social awareness, self-efficacy, and emotion regulation. Administer 1-3 times per year.

Referral System

Educators can refer students for academic support, counseling, behavior intervention, or family outreach. Status tracking from submission through resolution.

Smart Groups

Auto-generated student groups based on tier status, demographics, grade level, or custom filters. Target evidence-based interventions to the right students without manual sorting.

Leadership Dashboards

District and school-level views: chronic absenteeism rates, intervention effectiveness, behavior trends, strategic goal tracking. Disaggregate every report by demographics for equity analysis.

Strategy Library

Built-in playbook of research-based intervention strategies organized by domain and tier. Select strategies directly into intervention plans. Add your own institution-specific strategies.

Support Notes and Collaboration

Any educator can add timestamped notes to a student profile. Tag by category: academic concern, behavior, family contact, meeting note. Shared across the student's support team.

Behavior Incident Logging

Log incidents with type, severity, location, time, action taken, and follow-up. Track patterns over time. Percentage-based metrics give students room to improve.

Demographic Disaggregation

Every dashboard, report, and chart can be filtered by race/ethnicity, gender, ELL status, IEP/504, free/reduced lunch, grade level, and school. Built for equity-focused decision making.

Attendance Tracking

Daily attendance with chronic absenteeism alerts. Excused vs unexcused breakdown. Trend analysis by week, month, and term. Integrates with SIS data imports.

Configurable Thresholds

Every tier threshold is adjustable. Choose your grading system, set attendance cutoffs, define behavior incident thresholds, and customize SEL screener strength levels. Your district, your benchmarks.

In Development · April 2026

Compass: Your AI Assistant. On Your Server.

Compass is an AI assistant built for your district and trained by your teachers. It doesn't call OpenAI. It doesn't call Anthropic. It doesn't call Google. It runs entirely on the same server as the rest of Atlas K-12, powered by an open-source language model your district owns - and every time a teacher corrects an answer or marks one as exactly right, Compass learns how your district actually does MTSS. The longer you use it, the more it sounds like the people in your buildings.

Compass Agent

Not a generic chatbot. Compass is your district's AI - one that gets smarter every week from the corrections, thumbs-ups, and "no, here's how we actually do it" notes your teachers leave behind. Click the glowing compass icon from a student profile and it already knows who you're looking at. Strategy mode pulls from your curated library; Navigator mode answers "how do I" questions about the software.

  • Learns from your teachers: every correction shapes the next answer for everyone in the district
  • Knows your library: retrieval is scoped to the strategies and techniques your MTSS lead has approved
  • Knows the student: opens already aware of who you were just looking at, no re-typing context
  • Three power levels, your choice: swap the underlying model with a single config line - see Minimum System Requirements below for the full breakdown
  • Stays on-box: every model runs in-process via llama-cpp-python - no outbound API calls, ever, regardless of which tier you pick

District-Curated Technique Catalog

Atlas K-12 ships with a curated library of evidence-based pedagogical techniques drawn from IES, WWC, PBIS, CASEL, and NCII. Before your AI assistant suggests anything, your MTSS lead reviews every technique through a built-in wizard and decides which ones fit your district. Rejected techniques are filtered out of every recommendation, every draft, every retrieval.

  • Coverage across academics, behavior, attendance, social-emotional
  • Tier 1 / 2 / 3 fit and grade band tagging
  • Append-only audit log of every curation decision
  • District extension strategies route to a separate review queue
🔒

Six Layers of Network Isolation

Atlas K-12's main application process has zero outbound network access - enforced in code at the Python socket layer. A separate, sandboxed resource fetcher process is the only component permitted to make outbound requests, and its allowlist is restricted to eight public research hostnames.

  • Process separation enforced by socket guard
  • Code-level allowlist + OS firewall + egress proxy options
  • Circuit breaker stops crash loops at 10 violations / 60s
  • Verification test suite enforced on every release
Three Compass Tiers

Pick the Brain. Swap It Anytime.

Compass runs on a local open-source language model. Atlas K-12 supports three tiers out of the box, all Western-origin, all running entirely on hardware your district owns. Pick the one that matches your hardware today - and when your hardware grows or your district outgrows the smaller model, swap to the next tier with a single config line. No data migration, no retraining, no downtime beyond a server restart.

1

Compass Lite

Phi-3 Mini · Microsoft · 3.8B

Lightweight assistant for evaluation, single-school pilots, or older hardware. Answers are grounded and useful but reasoning is shallow. Runs on 8 GB RAM.

Recommended
2

Compass Standard

Mistral Small 3 24B · Mistral AI France · Apache 2.0

The sweet spot for most US districts. Strong reasoning, structured tool calls, refuses out of scope cleanly, asks clarifying questions when unsure. Runs on 32 GB RAM.

3

Compass Pro

Llama 3.3 70B · Meta US · Llama 3 License

For large districts or central servers. Near GPT-4 / Claude Sonnet quality at the local model tier. Best with a GPU for speed. Runs on 64-128 GB RAM.

What You Actually Get At Each Tier

Compass Lite
The Lightweight Option
What it gives you
  • Fast, simple answers from your strategy library
  • Software navigation help (where to find pages, how to log incidents)
  • Refuses cleanly anything outside MTSS scope
  • Phi-3 Mini was trained on curated educational content, so it punches above its weight on instructional questions specifically
  • Runs on hardware that's probably already in your building
What it doesn't
  • Limited multi-step reasoning ("if A then check B then suggest C")
  • Sometimes refuses ambiguous questions instead of asking a clarifier
  • Tends to be literal - doesn't catch nuance the way larger models do
Best for: Single-school pilots, MTSS team evaluations, early adopter teachers who want to try Compass before the district commits.
Recommended
Compass Standard
The Sweet Spot
What it gives you
  • Everything Compass Lite does, dramatically better
  • Multi-step reasoning - "first check tier history, then look at the data, then suggest the next step"
  • Asks clarifying questions when the educator's intent is ambiguous, instead of guessing
  • Structured tool calls - Compass can look up the student's context, query the catalog, and pull research before answering
  • Self-critique loop - drafts an answer, critiques it for accuracy, revises if anything's off
  • Specifically tuned by Mistral AI for instruction-following and refusal discipline
  • Runs on a standard 32 GB office desktop with no GPU required
Trade-offs
  • Won't match GPT-4 or Claude on the very most complex reasoning questions
  • Slightly slower than Lite (~3-5 seconds per answer on a CPU-only box)
Best for: About 95% of US districts. If your hardware can handle it (32 GB RAM), this is the right call. The reasoning quality jump from Lite is dramatic and the hardware cost is modest.
Compass Pro
The Flagship
What it gives you
  • Everything Compass Standard does, sharper and more nuanced
  • Near GPT-4 / Claude Sonnet quality on reasoning tasks - the gap to closed-source frontier models nearly closes at this tier
  • Handles multi-paragraph complex queries gracefully ("Marcus is failing math, his attendance has dropped, his behavior referrals are climbing - what do I do?")
  • Best refusal discipline of any local model - least likely to be tricked, jailbroken, or talked out of its scope
  • Most mature ecosystem - Llama is the most-tested, most-documented open-source model in production today
  • With a GPU (RTX 3090, 4090, A6000, etc.) the answers come back in under a second
Trade-offs
  • Needs 64 to 128 GB RAM - standard office desktops won't fit it
  • GPU strongly recommended for usable speed (without one, expect 10-20 seconds per answer)
  • Larger disk footprint (40 GB for the model file alone)
Best for: Large districts (5,000+ students), districts with central server infrastructure, organizations that want absolute flagship quality and have the hardware budget to back it.

Swap anytime. Every Compass tier uses the same prompt architecture, the same retrieval pipeline, the same correction-learning loop, and the same district-curated technique catalog. The only thing that changes between tiers is which open-source model file the runtime loads. Your data, your corrections, and your library carry over unchanged.

New Side by Side

Compare Us Against the Competition

Pick a vendor you are evaluating (Panorama, Branching Minds, eduCLIMBER, PowerSchool, Frontline, i-Ready) and see exactly where Atlas K-12 stacks up across 12 features. We are competitive on the work that matters at a fraction of the cost.

See the Comparison →
The District IT Director Pitch

“Atlas K-12's main application process has no outbound network access whatsoever, enforced at the Python socket layer. A separate fetcher process is the only component permitted outbound, restricted to eight public research hostnames by three independent mechanisms. Student data never leaves the server. No telemetry, no analytics, no hosted AI calls. The local AI assistant runs on-box and has no network access at all.”

That's the answer that gets you past procurement.
For Your IT Department

Minimum System Requirements

Atlas K-12 runs entirely on hardware your district owns. No cloud subscription, no monthly per-seat fee, no outbound data flows. Pick the tier that matches your district size and the AI horsepower you want behind Compass - you can upgrade later by swapping a single config line.

1

Demo / Pilot

Single building, under 500 students. Compass runs in lightweight mode. Good for evaluation, MTSS team pilots, or a single-school deployment.

  • CPU: Modern dual-core (Intel i3, AMD Ryzen 3, or Apple M1 and up)
  • RAM: 8 GB
  • Disk: 10 GB free
  • OS: Windows 10/11, macOS 12+, or Linux (Ubuntu 22.04+)
  • Network: None required at runtime
  • Compass model: Phi-3 Mini (Microsoft, 3.8B parameters)
  • Browser: Any modern Chrome, Edge, Firefox, or Safari
Recommended
2

Standard District

Multi-school district up to 5,000 students. Compass runs the recommended Mistral Small 3 24B model with full reasoning, tool calling, and self-critique. The sweet spot for most US districts.

  • CPU: Modern quad-core (Intel i5-10400 or better, AMD Ryzen 5 5600 or better, Apple M2 or better)
  • RAM: 32 GB
  • Disk: 30 GB free (SSD strongly recommended)
  • OS: Windows 11 Pro, macOS 13+, or Linux server distribution
  • Network: None required at runtime; one-time download for AI model
  • Compass model: Mistral Small 3 24B (Mistral AI, France, ~14 GB)
  • GPU: Optional (any NVIDIA card with 8+ GB VRAM accelerates Compass 2-3x)
  • Browser: Any modern Chrome, Edge, Firefox, or Safari
3

Enterprise

Large districts (5,000+ students), multi-tenant servers, or districts that want the largest available local model. Compass runs Llama 3.3 70B with GPU offload for near-instant answers at scale.

  • CPU: 8+ core server class (Intel Xeon, AMD EPYC, Threadripper, or equivalent)
  • RAM: 64 to 128 GB
  • Disk: 100 GB free NVMe SSD
  • OS: Linux server (Ubuntu 22.04 LTS or RHEL 9 recommended) or Windows Server 2022
  • Network: None required at runtime; admin VPN for remote management optional
  • Compass model: Llama 3.3 70B (Meta, US, ~40 GB)
  • GPU: NVIDIA with 24+ GB VRAM (RTX 3090, RTX 4090, A5000, A6000, or better) - cuts Compass response time by 5-10x
  • Browser: Any modern Chrome, Edge, Firefox, or Safari
Notes for IT
  • Zero ongoing internet dependency. After the one-time install and AI model download, Atlas K-12 runs without any outbound network access. Students and teachers use it inside your firewall.
  • Open-source AI models only. Compass uses local GGUF models (Phi-3 from Microsoft, Mistral Small 3 from Mistral AI in France, Llama from Meta). All Western-origin, all open-source, all run on hardware you own. No Chinese-origin models in any tier.
  • Single-server deployment. Atlas K-12 runs as a single Python process plus a sandboxed fetcher subprocess. No Kubernetes, no microservices, no container orchestration required.
  • SQLite by default. Built-in SQLite database is fine for districts up to ~10,000 students. Districts beyond that can swap to PostgreSQL with a config change.
  • Upgrade path. Moving from Tier 1 to Tier 2 or Tier 2 to Tier 3 is a one-line config change plus downloading the larger AI model. Your data and configuration stay intact.
  • RAM is the gating factor. The Compass AI model has to fit in RAM. CPU and GPU only affect speed; if the RAM isn't there, the larger model simply won't load.

Every Student Deserves the Same Shot

The schools with the biggest budgets get premium early warning systems, premium intervention tracking, and premium student support platforms.

Everyone else gets

endless spreadsheets

and a prayer.

That is not good enough. Every child deserves the same level of intervention, tracking, and support, regardless of what their district can afford. Atlas K-12 gives every school a premium student support platform without the premium price tag.

The Cadillac plan. Without the Cadillac price.

No Two Atlas Builds Are Alike

Every District Gets Their Own Custom Build

Atlas is not one platform shipped to every school. When your district signs on, we write the underlying code to fit your data, your systems, and your workflow. The features look the same. The dashboards look the same. The MTSS framework is the same. But the engine running underneath is built specifically for how your district operates. No two Atlas installations are identical.

🔨
Custom Data Layer
Your SIS, your grading scale, your attendance codes, your discipline taxonomy. We rewrite the import and transform logic to match your exact data shape.
Custom Workflow
Your tier thresholds, your intervention rounds, your screening calendar, your Care Team meeting cadence. The code is tuned to how your district actually runs MTSS, not how a vendor thinks it should be run.
📊
Custom Reporting
Your state report formats, your accreditation needs, your federal funding requirements, your board-level dashboards. We code the reports your district actually has to file.
Other Platforms
  • ✗ One codebase shipped to every district
  • ✗ You map your data to their schema
  • ✗ You retrain staff to fit their workflow
  • ✗ Reports come pre-built; you adapt
  • ✗ If your district is unique, you are out of luck
Atlas
  • ✓ Custom build per district, baked into the code
  • ✓ Your schema becomes the source of truth
  • ✓ Workflow code adapts to your existing process
  • ✓ Reports are coded to your exact specifications
  • ✓ If your district is unique, that's the whole point

Your district is one of a kind. Your platform should be too.

Actively Seeking Partners

We're Looking for Districts to Build This With

Atlas K-12 is in active development and we are seeking school districts willing to pilot the platform during the build process. Your feedback shapes the product. You get early access to every feature as it's built, direct input on what gets prioritized, and when development is complete, founding partner pricing that locks in permanently. This is your chance to help build the tool your students need.

Join the Early Access List Become a Development Partner

Ready to Navigate Outcomes?

Request a free 30-day demo with all features unlocked. No credit card required.

Request Demo Contact Sales Download Brochure (PDF)

[email protected]