Ditempa Bukan Diberi — Forged, Not Given
Live data from /health and check_vital. Auto-refresh every 10 seconds.
Last update: --
Install the world's first Constitutional AI Governance System in seconds:
pip install arifosmcp==2026.3.14npm install -g @arifos/mcp{
"mcpServers": {
"arifos": {
"command": "npx",
"args": ["-y", "@arifos/mcp"]
}
}
}arifOS is not a wrapper—it is a governed intelligence organism. It metabolizes information through a Double Helix of specialized organs across three engines:
Stages: 000, 111, 222, 333, 444
Role: Reason, sense, ground
Floors: F2 (Truth), F4 (Clarity), F7 (Humility), F8 (Genius)
Stages: 555, 666
Role: Empathy, memory, ethics
Floors: F5 (Peace²), F6 (Empathy), F9 (Anti-Hantu)
Stages: 777, 888, 999
Role: Forge, judge, seal
Floors: F1 (Amanah), F3 (Tri-Witness), F10-F13
The complete tool surface distributed across 5 sovereign layers:
Constitutional airlock—establish session jurisdiction
Legacy alias for init_anchor
Kill switch—invalidate governed session
Tool surface declaration at boot
Stage conductor—routes ΔΩΨ transitions
Full pipeline trigger—INIT → SEAL in one call
Governed reasoning under F2/F4/F7
Metacognitive integration—checks own output
Epistemic intake—grounds claims before reasoning
Structured evidence map across sources
Live web search—raw reality acquisition
URL → normalized evidence artifact
Adversarial critique—safety and maruah check
Consequence simulation before action
Code + env actions under F11 gate
Constitutional semantic recall from Vault
Tri-witness verdict engine
Output validation—ALLOW/HOLD/VOID
F1–F13 live inspection + scoring
F12 injection guard—prefilter all inputs
888_HOLD registry + human escalation bus
System health—ΔS, peace², Ω₀ telemetry
Live floor scores + pipeline trace
Commit decision + telemetry to ledger
Merkle integrity check—tamper detection
These are non-negotiable structural laws enforced at every stage:
≥ 0.5 — Reversibility / audit mandate
≥ 0.99 — Anti-hallucination
≥ 0.95 — Human·AI·Earth consensus
≤ 0 — Entropy reduction
≥ 1.0 — Non-destructive power
≥ 0.70 — Serves weakest stakeholder
0.03–0.20 — Gödel uncertainty band
≥ 0.80 — Internal coherence
< 0.30 — Anti-Hantu limit
LOCK — No consciousness claims
LOCK — Identity verification
< 0.85 — Block adversarial control
HUMAN — Human final authority
/health
System health check—returns version, status, capability map
/tools
List all 25 canonical tools with JSONSchema parameters
/dashboard
Live metabolic loop visualization
/mcp
MCP protocol endpoint (Streamable HTTP)
External Validation: ✅ PASSED (2026-03-14)