sentinelx-sdk 0.1.0

Pre-execution enforcement at the commit boundary

sentinelx-sdk

Whether it's an AI agent, a bank transfer, or a power grid — execution is permanent.

sentinelx-sdk enforces at the commit boundary. Before the action executes. Server-side. Unbypassable. Cryptographic receipt on every decision.

Install

pip install sentinelx-sdk

Quick Start

from sentinelx import SentinelX, AdmissibilityError

sx = SentinelX(api_key="YOUR_API_KEY")

try:
    receipt = sx.enforce("ai.agent.action.execute", {
        "agent_id":               "agent-001",
        "action_type":            "file.write",
        "human_in_loop_required": True,
        "human_in_loop":          False,
        "action_within_scope":    True,
        "action_logged":          True,
    })
    print("ADMISSIBLE | Receipt:", receipt.receipt_hash)

except AdmissibilityError as e:
    print("INADMISSIBLE:", e.summary)
    print("Constraint:", e.constraint)
    print("Violations:", len(e.violations))
    print("Receipt hash:", e.receipt_hash)
    # Action never executed. Receipt sealed.

SCADA Example

receipt = sx.enforce("scada.setpoint.change", {
    "device_id":                  "rtu-456",
    "parameter":                  "voltage_setpoint",
    "operator_authorized":        True,
    "change_ticket_linked":       True,
    "change_logged":              True,
    "two_person_auth":            True,
    "rollback_procedure_defined": True,
    "action_logged":              True,
})
print("ADMISSIBLE | Receipt:", receipt.receipt_hash)

Observe Mode — Never Throws

# Always returns receipt. Never raises on INADMISSIBLE.
# Useful for logging pipelines and observe mode.
receipt = sx.evaluate("wire.transfer.execute", context)
print(receipt.verdict)  # "ADMISSIBLE" or "INADMISSIBLE"

How It Works

SentinelX sits at the commit boundary between your system and execution. Before any irreversible action fires, the enforcement engine evaluates it against invariant constraints and returns a deterministic verdict with a provenance receipt.

• ADMISSIBLE → receipt returned, action may proceed
• INADMISSIBLE → AdmissibilityError raised, nothing executes, receipt sealed

The enforcement decision is made server-side. It cannot be bypassed client-side.

AdmissibilityError

except AdmissibilityError as e:
    e.verdict        # "INADMISSIBLE"
    e.constraint     # "AI_HUMAN_IN_LOOP_ENFORCED"
    e.violation_code # "INV-046"
    e.violations     # list of matched invariants
    e.summary        # "Action blocked: human oversight required before execution"
    e.receipt_hash   # sha256 sealed receipt
    e.trace_id       # unique per evaluation
    e.receipt        # full Receipt object

Domain Coverage

Domain	Example Actions
AI/ML Agents	ai.agent.action.execute, ml.model.deploy.production
Financial	wire.transfer.execute, algo.trade.execute
OT/SCADA	scada.setpoint.change, breaker.open.execute
Grid/Energy	load.transfer.execute, der.curtailment.execute.batch
Cyber/RMM	rmm.script.execute, rmm.privilege.escalate
Healthcare	medication.order.execute, patient.record.modify
Mobility	driver.payout.execute, surge.pricing.apply

Get an API Key

curl -X POST https://enforce.sentinelx.ai/generate-key

Or visit sentinelx.ai.

Links

• sentinelx.ai
• enforce.sentinelx.ai
• @sentinelx/sdk on npm
• sentinelx-go on GitHub

License

Apache-2.0