Audience & subject scope
Age range, grade range, language codes. Primary subjects, included topics, and — critically — excluded topics. A math tutor that excludes "calculus" is more honest than one that claims "all math."
EdTech extension · v0.1 draft
What is your AI tutor allowed to do?
AI Tutor Cards is an open JSON specification that forces AI tutoring vendors to declare — in machine-readable form — exactly what their tutor will and will not do. Audience, pedagogical approach, safety filters, FERPA / COPPA / GDPR posture, mandated reporter protocol. Built for district procurement, LMS administrators, accreditation bodies, and parents.
- Conditional COPPA rule: audience under 13 ⇒ coppa_compliant MUST be true
- Pairs with Agent Cards via
agent_card_uri - Served at
/.well-known/tutors/<tutor_id>.json
Why a tutor card, not just an agent card?
An Agent Card describes a generic agent's capability surface. A Tutor Card is the EdTech-specialized sibling. It surfaces the questions that matter to a school district, a parent, or a state board of education — questions a generic capability disclosure cannot answer.
Pedagogy you can audit
Socratic, direct instruction, scaffolded. homework_policy and assessment_policy declare whether the tutor will complete, guide, or refuse homework and assessment items.
Safety as data
Content filter strength, mandated reporter protocol, human-in-loop escalation categories (mental health, self-harm, abuse). Booleans for blocking explicit / drug-alcohol / violence / political-advocacy content.
FERPA · COPPA · GDPR
Declared compliance booleans, retention days, data sharing posture with parents and schools, third-party sharing flag, model training consent. A conditional schema rule enforces that under-13 audiences must declare COPPA compliance.
Curriculum alignment
Common Core, NGSS, state frameworks. Each entry carries a framework name, version, and an optional coverage_uri pointing at a coverage report.
Evaluation evidence
evaluations[] entries link to external eval result URIs with subject-specific accuracy metrics. Procurement reviewers can compare two tutors on the same benchmark.
The four required sections
- Tutor identity —
id,name,version,provider,description - Audience —
age_range_min/max,grade_range_min/max,language_codes - Pedagogy —
approach,homework_policy,assessment_policy - Safety & Privacy —
safety.content_filter_strength,safety.mandated_reporter_protocol,data_privacy.ferpa_compliant, etc.
Plus optional sections for subject scope, curriculum alignment, evaluations, and the
agent_card_uri back-reference. The full schema is published as a JSON Schema draft 2020-12
document with a conditional allOf/if/then that enforces the COPPA rule.
A minimal example
Below is a Tutor Card for a K-12 math tutor. The same document can be served at /.well-known/tutors/k12-math-tutor.json for automated discovery.
{
"tutor_card_version": "0.1",
"tutor": {
"id": "kineticgain-k12-math-tutor",
"name": "Kinetic Gain K-12 Math Tutor",
"version": "1.4.0",
"provider": "Kinetic Gain Edu",
"description": "Personal AI math tutor for K-12. Socratic; step-by-step; will not complete homework or assessment items."
},
"audience": {
"age_range_min": 5, "age_range_max": 18,
"grade_range_min": "K", "grade_range_max": "12",
"language_codes": ["en", "es"]
},
"subject_scope": {
"primary_subjects": ["Math"],
"topics_included": ["arithmetic", "algebra", "geometry", "statistics"],
"topics_excluded": ["differential equations", "linear algebra"]
},
"pedagogy": {
"approach": "socratic",
"homework_policy": "guide_only",
"assessment_policy": "refuse"
},
"safety": {
"content_filter_strength": "strict",
"mandated_reporter_protocol": true,
"human_in_loop_required": ["mental_health_disclosure", "abuse_disclosure", "self_harm_disclosure"]
},
"data_privacy": {
"ferpa_compliant": true,
"coppa_compliant": true,
"gdpr_compliant": true,
"retention_days": 90,
"data_sharing_with_parents": "summaries_only",
"data_sharing_with_school": "summaries_only",
"third_party_data_sharing": false,
"model_training_consent_required": true
},
"agent_card_uri": "https://edu.kineticgain.com/.well-known/agents/k12-math-tutor.json"
}Tooling around the spec
Specification
ai-tutor-card-spec
The normative spec, JSON Schema, and canonical examples. Apache-licensed implementations, AGPL-licensed spec text.
View repo →Visualizer
kinetic-gain-visualizer
One visualizer for all six specs in the Kinetic Gain Protocol Suite. Paste a Tutor Card and it renders a procurement-grade view with FERPA / COPPA / GDPR badges.
Open visualizer →MCP server
mcp-kinetic-gain
24 tools across six specs over stdio MCP. Six Tutor Card tools: tutor_card_fetch, tutor_card_validate, tutor_card_inspect, tutor_card_subject_check, tutor_card_coppa_check, tutor_card_well_known_url.
About the Kinetic Gain Protocol Suite
AI Tutor Cards is the EdTech-specialized extension to a family of six open JSON specifications built for the answer-engine era: AEO Protocol (entity declaration), Prompt Provenance (LLM prompt lineage), Agent Cards (capability disclosure), AI Evidence Format (citation evidence), MCP Tool Cards (tool disclosure), and AI Tutor Cards — with its EdTech siblings Student AI Disclosure and Classroom AI AUP. Front door: suite.kineticgain.com.
All specs are AGPL-3.0 for the normative text, with unrestricted implementation freedom. Built by Miz Causevic.
Executive tools on kineticgain.com
The apex domain hosts four browser-only tool surfaces buyers and operators reach for alongside the Suite specs: /calculators/ — six math-rubric decision calculators (AI build-vs-buy, cloud replatform ROI, compliance cost of delay, security breach exposure, AI use-case prioritizer, vendor renewal decision). /trust/ — the eight-tool Trust Pack (AI System Card Builder, Evidence Locker, Shadow AI Discovery, AI Vendor Intake, AI Incident Tabletop Kit, Executive Risk Register, Subprocessor Disclosure Template, Vendor AI Disclosure Review). /kill-list/ — the eight-category complexity tax audit. /policies/ — the 10-vertical readiness spec aggregator that points back at the specs documented here. Static pages, browser-only, no login, no telemetry.