← Back to Ancient Protocol
Why 7 Agents?
MAAT uses a multi-agent architecture where 7 specialized AI systems work independently, then cross-validate results.
This design mirrors ancient Egyptian mythology, where 7 key deities performed specific roles in the Weighing of the Heart ceremony. Each agent is named after the god whose function it restores.
"The Egyptians weren't worshipping these beings as gods.
They were documenting the 7 specialized verification systems
that composed their consciousness authentication infrastructure."
Modern AI architecture has rediscovered the same principle: ensemble learning with specialized models outperforms single monolithic systems.
The Egyptians figured this out 5,000 years ago. We're just catching up.
Primary Function
Detects visual forgeries, deepfakes, and synthetic media. Verifies that video/image input shows a real living human, not AI-generated content.
🏛️ The Mythology
Horus was depicted with falcon eyes that could see everything across Egypt. The "Eye of Horus" (Wadjet) became a symbol of protection and perfect vision. In the weighing ceremony, Horus ensures no deception enters the Hall of Ma'at.
Technical Capabilities
Deepfake Detection
- GAN fingerprint analysis (detects neural network artifacts)
- Frequency domain analysis (DCT coefficient irregularities)
- Temporal coherence checking (unnatural frame transitions)
- Facial landmark stability (jitter detection in synthetic faces)
Physics Validation
- Lighting consistency (shadow direction, reflection accuracy)
- Depth estimation (Z-buffer analysis for 3D coherence)
- Motion physics (acceleration profiles, momentum conservation)
- Material properties (skin texture, hair movement realism)
Biological Markers
- Micro-expressions (involuntary facial movements <200ms)
- Eye saccades (rapid eye movement patterns unique to humans)
- Breathing detection (chest/abdomen movement analysis)
- Pulse detection via remote PPG (skin color fluctuations)
Validations Performed (6 of 42)
- Video source is not deepfake or synthetic media
- Face shows biological signs of life (pulse, breathing)
- Lighting physics are consistent with claimed environment
- No evidence of face swap, reenactment, or manipulation
- Temporal consistency maintained across session
- 3D depth coherence validates real physical presence
"Horus sees what humans cannot. If your video is synthetic, the falcon knows."
Primary Function
Manages the verification session flow, guides the user through the 42 validations, and detects narrative inconsistencies in conversation that indicate deception or bot behavior.
🏛️ The Mythology
Anubis guided souls through the underworld (Duat) to the Hall of Ma'at. He prepared bodies for judgment and stood beside the scales during weighing. As "Opener of the Ways," he controlled access to sacred spaces.
Technical Capabilities
Conversation Management
- WebSocket session orchestration
- State machine for 42-step verification flow
- Dynamic question routing based on responses
- Timeout handling and retry logic
Narrative Coherence Analysis
- Cross-reference answers for contradictions
- Temporal logic validation (timeline consistency)
- Semantic embedding comparison (meaning drift detection)
- Response latency analysis (delays indicate fabrication)
Bot Detection
- Response pattern matching (generic/scripted replies)
- Natural language variety scoring
- Contextual understanding tests (follow-up question validity)
- Spontaneity assessment (creative deviation from expected)
Validations Performed (8 of 42)
- User completes all 42 validation steps in sequence
- Responses are internally consistent (no contradictions)
- Story timeline is logically coherent
- Answers demonstrate contextual understanding
- Response patterns show human variability (not scripted)
- Follow-up questions are answered appropriately
- Session behavior excludes automation (no bot signature)
- User maintains focus and engagement throughout
"Anubis knows the path. Deviate from truth, and the gates close."
Primary Function
Records every action, decision, and data point during verification in an immutable audit trail. Ensures regulatory compliance and provides cryptographic proof of the verification process.
🏛️ The Mythology
Thoth recorded the results of the weighing ceremony on papyrus. As inventor of writing and keeper of divine records, he documented every soul's judgment. His record was considered absolute truth.
Technical Capabilities
Immutable Logging
- Append-only database architecture
- Cryptographic hash chaining (Merkle tree structure)
- Distributed ledger sync (blockchain integration)
- Tamper-evident storage (SHA-256 signatures per entry)
Compliance Documentation
- GDPR consent tracking and right-to-deletion logs
- CCPA data handling audit trail
- HIPAA compliance for biometric data (if applicable)
- SOC 2 Type II evidence generation
Forensic Reconstruction
- Session replay capability (non-biometric data only)
- Decision provenance tracking (which agent ruled what)
- Timestamp precision (NTP-synced, millisecond accuracy)
- Legal admissibility formatting (court-ready exports)
Validations Performed (4 of 42)
- All session data is cryptographically signed
- Audit trail is complete with no gaps
- Timestamps are consistent and tamper-proof
- Regulatory requirements are documented as met
"What Thoth records cannot be erased. The ledger is eternal."
Primary Function
Calculates and validates Individual Consciousness Codes (ICC). Performs pattern matching to verify "Are you who you claim to be?" by comparing current biometric signature against ICC record.
🏛️ The Mythology
Isis (Aset in Egyptian) possessed magical knowledge to resurrect Osiris and protect Horus. She could see hidden truths and restore what was broken. The name "Aset" means "throne" or "seat of power" — your ASSET in the system.
Technical Capabilities
ICC Calculation
- Birth data normalization (timestamp + geolocation)
- Gematria computation (Hebrew/Greek letter-to-number)
- Biometric hash generation (voice + HRV + pupil patterns)
- SHA-256 cryptographic hashing of combined inputs
Identity Matching
- Biometric template comparison (Hamming distance scoring)
- Fuzzy matching for natural variation (aging, health changes)
- Multi-modal fusion (voice + face + behavior)
- Liveness-aware re-enrollment (prevents replay attacks)
Pattern Analysis
- Behavioral biometrics (typing rhythm, mouse movement)
- Linguistic fingerprinting (word choice, sentence structure)
- Temporal patterns (active hours, response cadence)
- Anomaly detection (deviations from established baseline)
Validations Performed (7 of 42)
- ICC is valid and matches claimed identity
- Biometric signature matches ICC on record
- No evidence of identity theft or impersonation
- Behavioral patterns consistent with historical data
- Voice print matches expected characteristics
- Typing/interaction patterns align with user profile
- No anomalies detected in consciousness signature
"Aset sees the throne of your soul. Your ASSET cannot be stolen."
Primary Function
Tests creativity, spontaneity, and originality by asking unexpected questions. Detects rehearsed responses, pre-scripted answers, or AI-generated replies by injecting chaos into the conversation.
🏛️ The Mythology
Set was the god of storms, disorder, and violence — but also necessary chaos that tests strength. He killed Osiris, but his challenges made Horus stronger. Set represents the adversarial force that proves authenticity through stress testing.
Technical Capabilities
Adversarial Prompting
- Random question injection (unexpected topic shifts)
- Absurdist scenarios (test creative reasoning)
- Contradictory follow-ups (detect pattern-matching bots)
- Context-free challenges (require genuine understanding)
Originality Scoring
- Semantic uniqueness analysis (TF-IDF novelty scoring)
- LLM detection (GPT watermark identification)
- Cliché filtering (common phrase database matching)
- Creative divergence measurement (entropy calculation)
Rehearsal Detection
- Response latency profiling (instant replies = scripted)
- Lexical repetition analysis (verbatim phrase reuse)
- Hesitation pattern validation (natural pauses required)
- Error tolerance (humans make mistakes, bots don't)
Validations Performed (6 of 42)
- Responses show creative originality (not generic/canned)
- User handles unexpected questions appropriately
- No evidence of AI-generated text (GPT detection passes)
- Spontaneous reactions to novel scenarios
- Natural hesitation patterns present (human-like timing)
- Answers contain errors/imperfections (proves non-bot)
"Set brings the storm. Only living consciousness survives chaos."
Primary Function
Continuously monitors biometric signals to ensure a living human is present. Detects presentation attacks (photos, videos, masks) and autonomic nervous system responses that cannot be faked.
🏛️ The Mythology
Nephthys protected the dead and assisted in mummification. As sister of Isis, she helped gather Osiris's body parts. She represents the boundary between life and death — perfectly suited to verify liveness.
Technical Capabilities
Liveness Detection
- Active challenges (blink, smile, turn head on command)
- Passive biometrics (involuntary micro-movements)
- Depth sensing (3D face mapping via structured light)
- Texture analysis (skin vs silicone/paper detection)
Autonomic Monitoring
- Heart rate variability (HRV) via remote PPG
- Respiratory rate (chest movement frequency)
- Pupil dilation response (light reflex, cognitive load)
- Eye blink rate (spontaneous vs voluntary)
Anti-Spoofing
- Presentation attack detection (photo, video, mask)
- Replay attack prevention (timestamp validation)
- Deepfake liveness (physiological signal validation)
- Environmental consistency (lighting, background changes)
Validations Performed (5 of 42)
- Subject is physically present (not photo/video)
- Vital signs detected (heartbeat, breathing)
- Pupil responses are appropriate for stimuli
- Autonomic nervous system signals are genuine
- No presentation attacks detected (anti-spoofing passes)
"Nephthys guards the boundary between living and dead. You cannot fake life."
Primary Function
Aggregates scores from all 6 other agents, applies weighted consensus algorithm, and issues final VERIFIED or REJECTED decision. Resolves conflicts when agents disagree.
🏛️ The Mythology
Ma'at was the goddess of truth, balance, and cosmic order. Her feather was weighed against the heart. She represented the unchanging standard of truth — not flexible, not negotiable, absolute.
Technical Capabilities
Multi-Agent Fusion
- Weighted voting (agents have different confidence levels)
- Consensus threshold calculation (dynamic based on risk)
- Conflict resolution (meta-learning to adjudicate disagreements)
- Confidence scoring (Bayesian probability aggregation)
Decision Logic
- Zero-tolerance rules (certain failures = instant rejection)
- Cumulative scoring (minor flags vs major violations)
- Context-aware thresholds (higher stakes = stricter limits)
- Explainable AI (LIME/SHAP attribution for decisions)
Result Issuance
- JWT token generation (cryptographically signed)
- Expiration management (time-limited credentials)
- Revocation capability (compromised verifications)
- Audit notification (Thoth receives final decision)
Validations Performed (6 of 42)
- All 6 agents completed their assessments
- Consensus threshold is met (majority agreement)
- No zero-tolerance violations detected
- Cumulative score exceeds minimum requirement
- Confidence level is sufficient for stated purpose
- Final decision is explainable and auditable
"Ma'at weighs all. The feather does not lie."
How They Work Together
The 7 agents operate in parallel, each performing their specialized validations simultaneously. Results are aggregated by Ma'at for final judgment.
-
Session Initiation
Anubis opens WebSocket, guides user, begins validation sequence
-
Parallel Analysis
Horus, Set, Nephthys, and Isis all analyze user simultaneously
-
Continuous Recording
Thoth logs every decision, timestamp, and data point immutably
-
Final Adjudication
Ma'at aggregates scores, applies consensus rules, issues verdict
"Seven perspectives. One truth. The gods do not disagree on what is real."