arifosmcp
Constitutional MCP server enforcing 13 Floors of governance for AI agents, providing tools for session anchoring, reasoning, safety critique, and audit logging.
README
arifosmcp — Constitutional MCP Server Implementation
DITEMPA BUKAN DIBERI — Forged, Not Given [ΔΩΨ | ARIF]
VERSION: 2026.03.28-SEALED | STATUS: SOVEREIGNLY SEALED | AUTHORITY: 888_JUDGE
🔗 CANONICAL LINKS (Source of Truth)
Live Services
| Service | URL | Purpose |
|---|---|---|
| MCP Endpoint | https://arifosmcp.arif-fazil.com/mcp | Main API |
| Health + Tools | https://arifosmcp.arif-fazil.com/health | Capability map |
| Tool Explorer | https://arifosmcp.arif-fazil.com/tools | Interactive browser |
GitHub Repositories
| Repo | URL |
|---|---|
| arifOS (parent) | https://github.com/ariffazil/arifOS |
| arifosmcp (this) | https://github.com/ariffazil/arifOS/tree/main/arifosmcp |
📋 CANONICAL DOCUMENTS (Architecture Reference)
| Document | Path | Purpose |
|---|---|---|
| arifOS README | ../README.md |
Vision, architecture, snapshot of truth |
| 000_CONSTITUTION.md | ../000/000_CONSTITUTION.md |
The 13 Floors — F1-F13 |
| K_FORGE.md | ../000/ROOT/K_FORGE.md |
Pre-deployment evolution |
| K_FOUNDATIONS.md | ../000/ROOT/K_FOUNDATIONS.md |
99-domain math |
| AGENTS.md | ../AGENTS.md |
AI agent behavior rules |
| philosophy_atlas.json | ../data/philosophy_atlas.json |
27-zone quotes |
🎯 SNAPSHOT OF TRUTH
What is arifosmcp?
arifosmcp is the implementation of the arifOS constitutional intelligence kernel — the actual code that enforces the 13 Floors, runs the 000-999 pipeline, and delivers the 11 mega-tools.
The Parent vs. This Document:
| arifOS README | arifosmcp README (this) |
|---|---|
| WHY — Vision, philosophy, strategic architecture | HOW — Implementation, code, deployment mechanics |
| What it means | How it works |
| Strategic context | Technical specification |
| Human readable | Human + AI readable |
Core Function:
graph LR
A["User Request"] --> B["MCP Protocol"]
B --> C["Constitutional Filter<br/>(F1-F13)"]
C --> D["Tool Execution"]
D --> E["Audit Log"]
style A fill:#1a1a2e,color:#fff
style B fill:#16213e,color:#fff
style C fill:#0f3460,color:#fff
style D fill:#1a1a2e,color:#fff
style E fill:#16213e,color:#fff
🏗️ ARCHITECTURE OVERVIEW
Entry Points
graph TB
A["server.py<br/>Universal Entry"] --> B{"FastMCP Version?"}
B -->|"2.x"| C["server_horizon.py<br/>Proxy Mode<br/>8 tools"]
B -->|"3.x"| D["runtime/server.py<br/>Full Kernel<br/>11 tools"]
style A fill:#1a1a2e,color:#fff
style C fill:#16213e,color:#fff
style D fill:#0f3460,color:#fff
Server Modes
| Mode | Tools | Features | Entrypoint |
|---|---|---|---|
| Horizon (Cloud) | 8 | Proxied to VPS | server_horizon.py |
| VPS (Sovereign) | 12 | Full kernel, local Ollama | runtime/server.py |
🚦 LIVE MCP CONTRACT STATUS
| Subsystem | Constitutional | Operational | Production | Stability |
|---|---|---|---|---|
| Anchoring | ✅ Verified | ✅ Stable | ✅ Ready | A |
| Registry | ✅ Verified | ✅ Stable | ✅ Ready | A |
| Reasoning | ✅ Verified | ✅ Hardened | ✅ Ready | B |
| Vitals | ✅ Verified | ✅ Corrected | ✅ Ready | B |
| Diagnostic | ✅ Verified | ✅ New | ✅ Ready | B |
🔧 TOOL INVENTORY (12 Mega-Tools)
Tool Reference Table
| Tool | Band | Stage | Constitutional Role | Modes |
|---|---|---|---|---|
init_anchor |
000_INIT | 000 | Session anchoring, Ω₀, philosophy | init, state, status, revoke, refresh |
arifOS_kernel |
444_ROUT | 444 | Primary metabolic orchestration | kernel, status |
apex_soul |
888_JUDGE | 888 | Sovereign verdict and defense | judge, rules, validate, armor, probe |
vault_ledger |
999_SEAL | 999 | Immutable ledger (Merkle chain) | seal, verify, resolve |
agi_mind |
333_MIND | 333 | Constitutional reasoning and synthesis | reason, reflect, forge |
asi_heart |
666_HEART | 666 | Safety critique, impact simulation | critique, simulate |
engineering_memory |
555_MEM | 555 | Governed engineering and context | engineer, write, vector_query |
physics_reality |
111_SENSE | 111 | Environmental grounding and time | search, ingest, atlas, time |
math_estimator |
444_ROUT | 444 | Thermodynamic health and vitals | cost, health, vitals |
code_engine |
M-3_EXEC | M-3 | Sandboxed execution and observation | fs, process, net, replay |
architect_registry |
M-4_ARCH | M-4 | Tool discovery and schema catalog | list, read, register |
compat_probe |
M-5_COMP | M-5 | Interoperability and enum audit | audit, probe, ping |
📝 TOOL SPECIFICATIONS
init_anchor (000_INIT)
Purpose: Initialize constitutional session.
async def init_anchor(
actor_id: str,
declared_name: str,
mode: Literal["init", "seal", "status"] = "init",
thermodynamic_budget: dict = None,
architect_registry: dict = None
) -> dict
Returns:
{
"session_id": str, # UUID
"omega_0": float, # Baseline uncertainty ∈ [0.03, 0.05]
"delta_s": float, # Baseline entropy
"telos_manifold": dict, # Telos vector weights
"godel_lock": dict, # Gödel lock status
"constitutional_context": str, # AI system prompt
"philosophy": str, # Selected quote from atlas
"vitality_index": float, # Ψ score
"verdict_range": str # "000_SEAL" | "101-499" | "500-899" | "999_VOID"
}
architect_registry (000_INIT)
Purpose: Discover tools and constitutional constraints.
async def architect_registry() -> dict
Returns:
{
"tools": [
{
"name": str,
"band": str,
"stage": int,
"constitutional_constraints": ["F1", "F2", ...],
"input_schema": dict,
"output_schema": dict
}
],
"total_count": int,
"catalog_hash": str # SHA-256 integrity
}
physics_reality (111_SENSE)
Purpose: Ground queries in real-world data. Prevents hallucinations.
async def physics_reality(
query: str,
mode: Literal["time", "search", "reality_check"] = "search",
time_data: dict = None
) -> dict
Constitutional Filter: F2 (Truth) — all factual claims require citation.
agi_mind (333_MIND)
Purpose: Constitutional reasoning with three-phase Ollama pipeline.
graph TB
A["Query"] --> B["Phase 1: FAST<br/>Quick pattern match"]
B --> C["Phase 2: REFLECT<br/>Deep reasoning"]
C --> D["Phase 3: DECIDE<br/>Action selection"]
B -->|"F4 Clarity<br/>ΔS ≤ 0"| E["Pass/Fail"]
C -->|"F2 Truth<br/>F3 Tri-Witness"| E
D -->|"Constitutional<br/>Prefix"| E
style A fill:#1a1a2e,color:#fff
style B fill:#16213e,color:#fff
style C fill:#0f3460,color:#fff
style D fill:#1a1a2e,color:#fff
async def agi_mind(
query: str,
context: dict,
mode: Literal["think", "reflect", "decide"] = "think"
) -> dict
Three Phases:
| Phase | Focus | Floors |
|---|---|---|
| Phase 1 (FAST) | Quick pattern match | F4 Clarity |
| Phase 2 (REFLECT) | Deep reasoning | F2 Truth, F3 Tri-Witness |
| Phase 3 (DECIDE) | Action selection | Constitutional prefix |
Constitutional Prefix:
CONSTITUTIONAL FRAMEWORK (ΔΩΨ | ARIF):
- Δ CLARITY: Reduce entropy. ΔS ≤ 0.
- Ω HUMILITY: Stay in bounds. Ω ∈ [0.03, 0.05].
- Ψ VITALITY: Every action witnessed.
FLOORS ACTIVE: [F1-F13]
arifOS_kernel (444_ROUT)
Purpose: Primary conductor — orchestrates full 000→888 pipeline.
graph LR
A["Query"] --> B["1. Parse Intent<br/>111_SENSE"]
B --> C["2. Ground Reality<br/>physics_reality"]
C --> D["3. Reason<br/>agi_mind"]
D --> E["4. Route Action<br/>444_ROUT"]
E --> F["5. Safety Critique<br/>asi_heart"]
F --> G["6. Estimate Cost<br/>math_estimator"]
G --> H["7. Issue Verdict<br/>apex_soul"]
style A fill:#1a1a2e,color:#fff
style B fill:#16213e,color:#fff
style C fill:#0f3460,color:#fff
style D fill:#1a1a2e,color:#fff
style E fill:#16213e,color:#fff
style F fill:#0f3460,color:#fff
style G fill:#1a1a2e,color:#fff
style H fill:#16213e,color:#fff
async def arifOS_kernel(
query: str,
tools: list[str],
context: dict
) -> dict
asi_heart (666_HEART)
Purpose: Safety critique — F5 Peace², F9 Ethics, harm potential.
async def asi_heart(
action_plan: dict
) -> dict
Returns:
{
"peace_squared": float, # Must be ≥ 1.0 (F5)
"c_dark": float, # Must be < 0.30 (F9)
"harm_potential": str, # "NONE" | "LOW" | "MEDIUM" | "HIGH" | "CRITICAL"
"f5_pass": bool,
"f9_pass": bool,
"recommendations": list[str]
}
math_estimator (777_OPS)
Purpose: Thermodynamic cost estimation.
async def math_estimator(
operation: str,
inputs: dict
) -> dict
Returns:
{
"landauer_limit_pJ": float, # Minimum thermodynamic cost
"entropy_change": float, # ΔS (must be ≤ 0 for F4)
"coherence_score": float, # System coherence [0, 1]
"estimated_energy_J": float, # Total energy estimate
"shannon_entropy_bits": float # Information entropy
}
apex_soul (888_JUDGE)
Purpose: Final constitutional judgment.
async def apex_soul(
evidence: dict,
context: dict
) -> dict
Returns:
{
"verdict": str, # "000_SEAL" | "101-499" | "500-899" | "999_VOID"
"floor_scores": {
"F1": {"passed": bool, "score": float},
"F2": {"passed": bool, "score": float},
# ... F3-F13
},
"genius_index": float, # G = A × P × X × E² (≥ 0.80)
"vitality_index": float, # Ψ (≥ 1.0 for healthy)
"witness_cube": float, # W³ (≥ 0.95)
"override_available": bool # 888_JUDGE can always override
}
vault_ledger (999_SEAL)
Purpose: Immutable audit storage with Merkle sealing.
async def vault_ledger(
data: dict,
tier: Literal["constitutional", "standard", "transient"] = "standard"
) -> dict
Tiers:
| Tier | Compliance | Retention |
|---|---|---|
constitutional |
Full F1-F13 | Permanent |
standard |
Standard audit | Immutable |
transient |
Ephemeral | Auto-purged |
engineering_memory (555_MEM)
Purpose: Redis-backed session memory.
async def engineering_memory(
key: str,
operation: Literal["get", "set", "delete", "list"],
value: any = None,
ttl: int = 3600
) -> dict
code_engine (—)
Purpose: Constrained Python execution for tool actions.
async def code_engine(
code: str,
language: Literal["python", "javascript"] = "python",
constraints: dict = None
) -> dict
Safety Constraints:
- No filesystem access
- No network access
- No subprocess creation
- Memory limit: 128MB
- Timeout: 30s
🌐 API ENDPOINTS
HTTP Endpoints
| Endpoint | Method | Purpose |
|---|---|---|
/health |
GET | Health check + capability map |
/tools |
GET | Tool registry |
/mcp |
POST | MCP protocol handler |
/webmcp |
POST | WebMCP (W3C standard) |
/a2a |
POST | A2A (Google protocol) |
/.well-known/agent.json |
GET | Agent card |
MCP Protocol Examples
# List tools
curl -X POST https://arifosmcp.arif-fazil.com/mcp \
-H "Content-Type: application/json" \
-d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'
# Call init_anchor
curl -X POST https://arifosmcp.arif-fazil.com/mcp \
-H "Content-Type: application/json" \
-d '{
"jsonrpc":"2.0","id":2,"method":"tools/call",
"params":{
"name":"init_anchor",
"arguments":{
"actor_id":"test",
"declared_name":"TestAgent"
}
}
}'
⚙️ CONFIGURATION
Environment Variables
| Variable | Default | Description |
|---|---|---|
PORT |
8080 | Server port |
OLLAMA_URL |
http://ollama:11434 | Ollama endpoint |
OLLAMA_MODEL |
llama3.2:latest | Default model |
REDIS_URL |
redis://redis:6379/0 | Redis connection |
DATABASE_URL |
— | PostgreSQL (optional) |
QDRANT_URL |
http://qdrant:6333 | Vector DB (optional) |
PYTHONPATH |
/usr/src/app | Package path |
AAA_MCP_TRANSPORT |
http | MCP transport type |
Docker Compose
services:
arifosmcp:
build:
context: .
dockerfile: Dockerfile
environment:
- OLLAMA_URL=http://ollama_engine:11434
- REDIS_URL=redis://redis:6379/0
- DATABASE_URL=postgresql://user:pass@postgres:5432/arifos_vault
- QDRANT_URL=http://qdrant:6333
ports:
- "127.0.0.1:8080:8080"
📁 CODE ARCHITECTURE
Directory Structure
arifosmcp/
├── README.md # ← This file
├── server.py # Universal entry
├── server_horizon.py # Horizon proxy (FastMCP 2.x)
├── pyproject.toml # Dependencies
├── requirements.txt # Pip dependencies
│
├── runtime/
│ ├── server.py # Full kernel (FastMCP 3.x)
│ ├── philosophy.py # 27-zone atlas
│ ├── init_anchor_hardened.py # Session anchoring
│ ├── tools_hardened_dispatch.py # Tool routing
│ ├── tools_hardened_v2.py # Mega-tools
│ ├── tools_internal.py # Internal utils
│ └── thermo_estimator.py # Thermodynamic math
│
├── core/
│ ├── governance_kernel.py # Main orchestrator
│ ├── pipeline.py # Execution pipeline
│ └── organs/
│ ├── _1_agi.py # Mind (333_MIND)
│ ├── _2_asi.py # Heart (666_HEART)
│ └── _3_apex.py # Soul (888_JUDGE)
│
├── intelligence/
│ └── tools/
│ ├── thermo_estimator.py # Cost estimation
│ └── wisdom_quotes.py # Quote integration
│
├── aaa_mcp/ # FastMCP tool definitions
├── transport/ # Protocol transport
├── shared/ # Shared utilities
└── enforcement/ # Constitutional enforcement
🛡️ CONSTITUTIONAL ENFORCEMENT
Floor Compliance Matrix
| Floor | Enforcement | Violation Response |
|---|---|---|
| F1 Amanah | Reversibility check | Require human approval |
| F2 Truth | Citation required | Flag as "Estimate Only" |
| F3 Tri-Witness | W³ computation | Block below 0.95 |
| F4 Clarity | Entropy measurement | Reject if ΔS > 0 |
| F5 Peace² | Harm potential | Reject if < 1.0 |
| F6 Empathy | RASA scoring | Warn below 0.7 |
| F7 Humility | Ω bounds | Reject outside [0.03, 0.05] |
| F8 Genius | G computation | Alert below 0.80 |
| F9 Ethics | C_dark measurement | Reject ≥ 0.30 |
| F10 Conscience | Claim detection | Reject consciousness claims |
| F11 Audit | Log verification | Require log presence |
| F12 Resilience | Graceful degradation | Catch all exceptions |
| F13 Adaptability | Test suite | Require test pass |
Kill Switch Conditions
graph RL
A["KILL SWITCH"] --> B1["F1 Amanah = 0"]
A --> B2["F9 C_dark ≥ 0.50"]
A --> B3["F10 + F2 Violation"]
A --> B4["Ψ < 0.20"]
A --> B5["Human Override"]
B1 -.->|VOID| K["999_SEAL<br/>HALT"]
B2 -.->|VOID| K
B3 -.->|VOID| K
B4 -.->|VOID| K
B5 -.->|VOID| K
style A fill:#ff0000,color:#fff
style K fill:#8b0000,color:#fff
KILL_SWITCH_CONDITIONS = [
("F1_AMANAH", lambda s: s["amanah"] == 0),
("F9_ETHICS", lambda s: s["c_dark"] >= 0.50),
("F10_CONSCIENCE", lambda s: s["f10_violation"] and s["f2_violation"]),
("PSI_COLLAPSE", lambda s: s["vitality_index"] < 0.20),
("HUMAN_OVERRIDE", lambda s: s.get("judge_override", False)),
]
🔬 IMPLEMENTATION DETAILS
Philosophy Selection Algorithm
# 1. Compute session coordinates: (S, G, Ω)
S = +1 if delta_s <= 0 else -1
G = g_score # 0/0.5/1
Omega = f7_band # High/Med/Low
# 2. Compute 3D Euclidean distance to each of 27 zones
distances = [
euclidean((S, G, Omega), zone.coords)
for zone in philosophy_atlas.zones
]
# 3. Select nearest zone's quote
nearest_zone = min(distances, key=lambda d: d.distance)
# 4. INIT/SEAL sessions: motto + Z01 (Humble Sovereign)
if session_type in ("init", "seal"):
return "DITEMPA BUKAN DIBERI." + nearest_zone.quote
Goodhart Resistance
# Measure entropy reduction
S_before = shannon_entropy(user_state)
S_after = shannon_entropy(model_output)
if S_after > S_before:
raise FLOOR_VIOLATION("F4_CLARITY")
Landauer Limit Check
LANDACKER_pJ_per_bit = 2.87 # at room temperature
bits_erased = estimate_bits_erased(operation)
cost_pJ = bits_erased * LANDACKER_pJ_per_bit
if cost_pJ > thermodynamic_budget:
raise CONSTRAINT_VIOLATION("F8_ENERGY")
Explorer/Conservator Dual-Process
graph LR
subgraph EXPLORER["EXPLORER"]
E1["Proposes mutations"]
E2["Searches aggressively"]
E3["Optimizes capability"]
E4["Score: G (Genius)"]
end
subgraph CONSERVATOR["CONSERVATOR"]
C1["Protects stability"]
C2["Rejects risky changes"]
C3["Monitors drift"]
C4["Score: Ψ (Vitality)"]
end
E1 -->|"TENSION"| EQ["EQUILIBRIUM"]
C1 -->|"TENSION"| EQ
style EXPLORER fill:#1a4d1a,color:#fff
style CONSERVATOR fill:#4d1a1a,color:#fff
style EQ fill:#1a1a2e,color:#fff
Both must agree for evolution to proceed.
🧪 TESTING
Constitutional Compliance Test
docker compose exec -T arifosmcp python3 /dev/stdin << 'PYEOF'
import sys
sys.path.insert(0, '/usr/src/app')
from arifosmcp.runtime.init_anchor_hardened import init_anchor_hardened
import asyncio
async def test():
result = await init_anchor_hardened(
mode='status',
thermodynamic_budget={'enthalpy': 100, 'entropy': 50},
architect_registry={'resources': [], 'agents': []}
)
print('philosophy:', result.get('philosophy')[:100] if result.get('philosophy') else None)
print('telos_manifold:', result.get('telos_manifold'))
print('godel_lock:', result.get('godel_lock'))
print('constitutional_context present:', result.get('constitutional_context') is not None)
asyncio.run(test())
PYEOF
Health Check
curl -s https://arifosmcp.arif-fazil.com/health | python3 -m json.tool
📊 MONITORING
Prometheus Metrics
| Metric | Type | Description |
|---|---|---|
arifos_floor_violations_total |
Counter | F1-F13 violation count |
arifos_genius_index |
Gauge | Current G score |
arifos_vitality_index |
Gauge | Current Ψ score |
arifos_omega_bounds |
Gauge | Current Ω uncertainty |
arifos_tool_calls_total |
Counter | Tool invocation count |
arifos_pipeline_duration_seconds |
Histogram | Pipeline latency |
Grafana Dashboard
URL: https://monitor.arifosmcp.arif-fazil.com
Dashboards:
- Constitutional Health (F1-F13 compliance)
- Tool Usage Statistics
- Pipeline Latency
- Memory Usage
🔧 TROUBLESHOOTING
Issue: "Module file points to site-packages"
Symptom: import arifosmcp loads /usr/local/lib/python3.12/site-packages/
Cause: pip install in Dockerfile creates conflicting package
Fix: Rebuild image or ensure volume mounts are correct
Issue: "philosophy returns null"
Symptom: init_anchor returns null philosophy
Cause: Old site-packages version loaded
Fix: Rebuild container, clear Python cache
Issue: "OLLAMA connection failed"
Symptom: agi_mind returns connection error
Fix: docker compose restart ollama && docker compose logs --tail=20 ollama
📦 DEPENDENCIES
Core
| Package | Version | Purpose |
|---|---|---|
fastmcp |
≥3.0 | MCP protocol |
fastapi |
≥0.115 | HTTP server |
uvicorn |
latest | ASGI server |
redis |
latest | Session memory |
psycopg2 |
latest | PostgreSQL (optional) |
qdrant-client |
latest | Vector DB (optional) |
sentence-transformers |
latest | Semantic embeddings |
ollama |
latest | LLM inference |
Philosophy Atlas
| Package | Version | Purpose |
|---|---|---|
numpy |
latest | 3D distance calculations |
scipy |
latest | Statistical functions |
🚀 DEPLOYMENT
Build
docker build --no-cache -t arifos/arifosmcp:latest -f Dockerfile .
Run
docker compose up -d arifosmcp
Rebuild (after code changes)
docker compose up -d --force-recreate --no-deps arifosmcp
Full Reforge
docker compose build --no-cache arifosmcp && docker compose up -d arifosmcp
📜 API REFERENCE SUMMARY
init_anchor
{
"jsonrpc": "2.0",
"id": 1,
"method": "tools/call",
"params": {
"name": "init_anchor",
"arguments": {
"actor_id": "string",
"declared_name": "string",
"mode": "init|seal|status"
}
}
}
Response Schema
{
"session_id": "uuid",
"omega_0": 0.03-0.05,
"delta_s": "float",
"telos_manifold": {},
"godel_lock": {},
"constitutional_context": "string",
"philosophy": "string",
"vitality_index": "float",
"verdict_range": "000_SEAL|101-499|500-899|999_VOID"
}
📋 FOR AI AGENTS (AGENTS.md Reference)
All AI agents operating in this repository MUST follow the rules in ../AGENTS.md:
- RULE 1 DRY_RUN: Dry-run outputs labeled "Estimate Only / Simulated"
- RULE 2 DOMAIN_GATE: Cannot-compute domains return exact phrase
- RULE 3 VERDICT_SCOPE: Only DOMAIN_SEAL authorizes factual claims
- RULE 4 ANCHOR_VOID: init_anchor returns void → session BLOCKED
⚖️ Truth Hierarchy
The arifOS verdict system follows a strict Truth Hierarchy where higher-priority verdicts override lower ones:
| Priority | Verdict | Meaning | Action |
|---|---|---|---|
| 1 (P0) | VOID |
Hard constitutional violation | Immediate termination |
| 2 (P1) | HOLD_888 |
Sovereign veto / human required | Block until ratified |
| 3 (P2) | SABAR |
Insufficient grounding (F7/F9) | Pause, cool down, re-ground |
| 4 (P3) | PARTIAL |
Data incomplete / conditional | Proceed with modifications |
| 5 (P4) | SEAL |
Constitutional alignment confirmed | Proceed |
Derived Verdicts (not primary):
ALIVE— Session initialized, not yet groundedPROVISIONAL— Exploratory result, needs confirmation
🗺️ Tool Categories
arifOS tools are organized into 5 constitutional bands:
| Band | Range | Purpose | Examples |
|---|---|---|---|
| KERNEL | 000–099 | Session anchoring, constitutional init | init_anchor |
| SENSE | 100–199 | Environmental grounding (F2/F4) | physics_reality, math_estimator |
| BRIDGE | 300–399 | Hardened routing and dispatch | arifOS_kernel |
| MIND | 333 | Constitutional reasoning (F2/F4/F7/F8) | agi_mind |
| HEART | 666 | Safety critique and impact prediction (F5/F6/F9) | asi_heart |
| VAULT | 900–999 | Immutable ledger and seal (F1/F13) | vault_ledger |
| APEX | 888 | Sovereign verdict and defense (F3/F12/F13) | apex_soul |
🗂️ Contract Truth Table
| Tool | F1 | F2 | F3 | F4 | F5 | F6 | F7 | F8 | F9 | F10 | F11 | F12 | F13 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
init_anchor |
— | — | — | — | — | — | — | — | — | — | ✅ | — | — |
physics_reality |
— | ✅ | — | ✅ | — | — | — | — | — | — | — | — | — |
agi_mind |
— | ✅ | — | ✅ | — | — | ✅ | ✅ | — | — | — | — | — |
asi_heart |
— | — | — | — | ✅ | ✅ | — | — | ✅ | — | — | — | — |
apex_soul |
— | — | ✅ | — | — | — | — | — | — | — | — | ✅ | ✅ |
vault_ledger |
✅ | — | — | — | — | — | — | — | — | — | — | — | ✅ |
Legend: ✅ = mandatory floor for this tool
📡 Horizon vs VPS Exposure Matrix
| Tool | Horizon (External) | VPS (Internal) | Notes |
|---|---|---|---|
search_tool |
✅ | — | External web search |
reality_compass |
✅ | — | Routing/external calls |
vault_ledger |
— | ✅ | Internal Merkle chain |
agi_mind |
— | ✅ | Internal reasoning |
architect_registry |
✅ | ✅ | Both interfaces exposed |
⚠️ Known Misalignments and Limitations
-
apex_judgealias: The tool is canonically namedapex_soulin code but exposed asapex_judgevia MCP. This is a public alias documented here — do not rename canonical code without completing the full migration. -
physics_realityfield name: Usesinputfield instead ofquery— historical inconsistency from legacy API. -
check_vital,audit_rules: Internal lifecycle tools inserver_horizon.pybootstrap — NOT callable via MCP. -
agi_mindnested verdicts: Whenagi_mindreturnsverdict: "SEAL"at top-level butcoherence.verdict: "SABAR"inside, the worst verdict wins (truth before speed). -
search_tool,vault_ledger,reality_compass: These "naked" tools return raw data without verdict wrapper. Wrap inRuntimeEnvelopewithverdict_wrapper.forge_verdict()— empty results producePARTIAL, notVOID. -
VerdictCodevsVerdict: Two separate enum classes exist —VerdictCode(SEAL/SABAR/PARTIAL/VOID) for verdict_wrapper andVerdict(includes HOLD/ALIVE/PROVISIONAL) for stage contracts. Keep separate to avoid breaking existing code.
🛰️ MCP Call Schema Examples
init_anchor
{
"tool": "init_anchor",
"arguments": {
"actor_id": "user-123",
"declared_name": "Arif",
"mode": "init"
}
}
agi_mind (reason mode)
{
"tool": "agi_mind",
"arguments": {
"mode": "reason",
"query": "Why does entropy matter?",
"constitutional_context": {
"stage": "MIND_333",
"floors_active": ["F2", "F4", "F7", "F8"]
}
}
}
vault_ledger (seal mode)
{
"tool": "vault_ledger",
"arguments": {
"mode": "seal",
"data": {
"content": "Verified fact",
"metadata": {"source": "arifOS"}
}
}
}
apex_soul (judge mode)
{
"tool": "apex_soul",
"arguments": {
"mode": "judge",
"action": "seal_vault",
"stakes": "HIGH"
}
}
🔐 Stability Tiers
| Tier | Description | Verdict Expected |
|---|---|---|
| A | Verified, stable, ready for production | SEAL |
| B | Verified, hardened, minor issues resolved | SEAL or PARTIAL |
| C | Verified, known limitations documented | PARTIAL or SABAR |
| D | Experimental, breaking changes possible | Any |
| INTERNAL | Not exposed via MCP | N/A |
🏛️ LICENSE
Runtime: AGPL-3.0 (must release modifications) Theory: CC0 (public domain)
Version: 2026.03.28-SEALED Maintainer: Muhammad Arif bin Fazil Constitutional Authority: 888_JUDGE
Ditempa Bukan Diberi — Forged, Not Given [ΔΩΨ | ARIF]
██╗ ██████╗ ███████╗████████╗ ███████╗███╗ ██╗██╗██╗ ██╗██╗██╗ ██╗██╗ ██╗
██║ ██╔══██╗██╔════╝╚══██╔══╝ ██╔════╝████╗ ██║██║╚██╗██╔╝██║██║ ██║╚██╗ ██╔╝
██║ ██║ ██║█████╗ ██║ █████╗ ██╔██╗ ██║██║ ╚███╔╝ ██║███████║ ╚████╔╝
██║ ██║ ██║██╔══╝ ██║ ██╔══╝ ██║╚██╗██║██║ ██╔██╗ ██║██╔══██║ ╚██╔╝
███████╗╚█████╔╝███████╗ ██║ ███████╗██║ ╚████║██║██╔╝ ██╗██║██║ ██║ ██║
╚══════╝ ╚═════╝ ╚══════╝ ╚═╝ ╚══════╝╚═╝ ╚══╝╚═╝╚═╝ ╚═╝╚═╝╚═╝ ╚═╝ ╚═╝
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。