Deterministic, tamper-evident accountability inside the IDE.

ARC XT operates at the moment of change, enforcing governed code boundaries before decisions ever reach pull requests, CI, or production.

Local-only governance Descriptive-only review Non-authorizing surface

Install ARC XT See the Decision Engine

Heuristic Evaluation

The 4-Tier Decision Engine

ALLOW

Low-risk changes like standard UI components or unit tests. Auto-validated and save proceeds normally.

WARN

Medium-risk edits like minor config changes. Requires explicit operator acknowledgment before proceeding.

REQUIRE_PLAN

High-risk logic modifications. Forces the operator to link a structured, pre-approved Blueprint Proof artifact.

BLOCK

Critical policy violations or unauthorized secure zone modifications. Save event is strictly blocked.

Cryptographic Audit Trail (audit.jsonl)

HASH-CHAINED RECORD

[REQUIRE_PLAN] Auth Module Mod

Blueprint linked: BP-920. Bounded authorization verified.

sha256: 8f4e2b0c9d...a1b2c3d4e5f6

[BLOCK] Credential Leak Prevented

Fail-closed triggered on auth_service.py:142

sha256: 3a1c9f8e7d...f1e2d3c4b5a6

Cryptographic Proof of Intent

data_object

Decision Record

Read-only descriptive reviews captured deterministically at save-time, never reconstructed after the fact.

assignment_turned_in

Blueprint Proof

Ensures high-risk changes are strictly mapped to instructional guidance through bounded authorization chains.

lock

Hash-chain Integrity

Tamper-evident audit events are mathematically chained so integrity remains provable over time.

ARC XT is not a coding assistant.

AI Coding Assistants	ARC XT Governance Layer
Generate code and suggest fixes	Verify deterministic boundaries
Maximize velocity and output	Enforce trust and operational accountability
Run as active sidecars	Run as a read-only observability substrate