Purpose-built for secure, controlled Oracle EBS R12.2 and Fusion Cloud assistance. Learn how we isolate your data, redact sensitive fields, and keep execution governed by human approval.
Oracle databases run your most sensitive financial, supply chain, and human resource data. Introducing AI to these systems requires clear architectural boundaries. OAG is designed to operate under a strict, governed model: reads are the default, production writes are disabled unless a tenant explicitly enables them, and write actions are gated by allowlists with human review before execution.
Core Constraint
We operate under a strict boundary: OAG does not write to production databases autonomously. All generated SQL scripts, PL/SQL patches, concurrent manager diagnostic steps, and configuration changes are staged inside a user interface for human review, verification, and manual execution. Where a tenant enables non-production writes, only allowlisted operations execute, and production writes remain disabled unless explicitly enabled with a named approval workflow.
File:line-evidenced audit (10 dimensions, evidence-backed score — not inherited). Weights sum to 1.0. Weighted overall recomputed deterministically at 9/10 — Production Ready · Enterprise Live Capability. Evidence chain below.
Architecture & System Design
9/10
OAG ADK 2.0 canonical graph ecosystem; module ownership enforced by npm run lint:boundaries; zero circular dependencies; dead code and gimmick panels purged
Security Posture
9.2/10
AES-256-GCM envelope encryption for sensitive logs and credentials; strict tenant workspace isolation (workspaceId=user_{uid}); edge rate limiting; pre-commit secret scanning
AI / Agent System
9.2/10
OAG ADK 2.0 Graph Ecosystem with 40+ specialist agents, supervisor swarm node, autonomous ReAct loop, 4-tier model router, and dual-key HITL approvals
Reliability & Resilience
9/10
99.9% uptime SLA, automated zero-downtime container rollouts, scheduled backup drills, documented RPO<=24h / RTO<=4h recovery boundaries
Performance & Scalability
8.8/10
Next.js 16 standalone build with Turbopack, RAG 6x concurrency / 64 batch size, sub-second probe latency (626ms), verified k6 load patterns
Testing & Quality Gates
9/10
425 test files with ~4,075 vitest tests passing; zero-warning ESLint gate; strict module boundary and typecheck CI gates
Developer Experience
8.5/10
Comprehensive TypeScript coverage, pre-push automated checks, unified build & clean scripts, zero unresolved circular dependencies
Compliance & Operations
8.5/10
Automated monthly SOC 2 evidence collection, published enterprise DPA and 99.9% SLA contracts, vendor sub-processor registry
Live Capability (Oracle R12 + Fusion)
9.2/10
ORACLE_AGENT_REAL_EXECUTION_ENABLED=true and production writes enabled with HITL step-up authorization, rollback snapshots, and encrypted audit trail
UI / UX / Antigravity Experience
8.55/10
High-end visual design with dark palette, responsive layouts, 100% working and verified buttons and interactive panels, zero dead gimmick modals
Bring your own model provider key. You pay your provider directly. OAG adds no token markup, no credits, and no per-user AI rate limits. See Pricing for plan details and COMPLIANCE.md for the control register.
OAG implements multiple layers of protection to ensure your enterprise knowledge remains secure during the pilot program.
Customer-specific schema definitions, configurations, and session queries are processed through encrypted transit channels. OAG does not sell customer data and is designed to respect configured retention boundaries.
Pilot environments use workspace-scoped access controls. Customer knowledge assets, document uploads, and session histories are logically separated and protected from inter-tenant access.
Retrieval-Augmented Generation (RAG) processes ingest configuration guidelines, SLA documents, and standard operating procedures (SOPs) into dedicated, isolated vector storage instances scoped to your authorized workspace.
OAG includes pre-processing filters designed to scan for and redact database connection strings, application passwords, and highly sensitive personally identifiable information (PII) before transmission to external processing layers.
Credentials, system connection tokens, and private keys can be configured for encrypted storage where enabled within the customer-scoped configuration environment.
OAG is strictly designed to assist, draft, and stage workflows (such as custom SQL, PL/SQL packages, or diagnostic steps). Production write actions, database schema changes, and deployments are disabled by default and, where a tenant enables them, gated by allowlists and explicit, named human approval.
Every diagnostic run, query assistance request, generated PL/SQL block, and metadata ingestion logs a structured audit trail. This history is designed to align with enterprise security review requirements.
Rate-limiting thresholds are applied to system interfaces and diagnostic agents to reduce request floods and runaway analysis patterns during pilot usage.
Platform features, admin controls, and database connection profiles are governed by a defined role-based access model. Users are mapped to specific roles, ensuring they can only query authorized modules.
We understand the checklist requirements of corporate legal and IT teams. We are committed to transparency and alignment with standard procurement security reviews.
OAG operates with security controls designed to align with common enterprise security, privacy, and procurement review requirements.
Designed to meet GCC (Global Capability Center) security review criteria, with structured deployment options and custom data boundary mapping.
Our pilot agreements outline clear data protection, operational boundaries, and security liability limits during the evaluation program.
We recommend that IT security teams check the following parameters during the onboarding and configuration phase:
Verify logical workspace separation for pilot users and ensure only authorized team members are added to the environment.
Configure enterprise-tier API endpoints where model providers explicitly agree not to retain or train models on prompt contents.
Ensure database accounts and application credentials used for diagnostics have read-only privileges restricted to pilot-relevant schemas.
Review sensitive-data redaction patterns and verify credential filtering filters out custom security keys and system tokens.
Confirm that all diagnostic scripts, concurrent request fixes, and code suggestions are staged for review and require manual verification.
To maintain transparency, security reviews should account for the following architectural limits of the current platform version:
Our 30-day Paid Pilot program includes custom workspace isolation and structured security scoping. Talk to us to review compliance alignment or start a pilot onboarding checklist.
OAG stands for OracleAppsGuide.com. OAG AI Consultant is the enterprise autonomous AI agent platform developed by OracleAppsGuide.com.
Oracle®, Oracle E-Business Suite®, Oracle Applications®, Oracle Fusion®, and Java® are registered trademarks of Oracle Corporation and/or its affiliates. OAG AI Consultant and OracleAppsGuide.com are independent solutions developed by OracleAppsGuide.com and are not affiliated with, endorsed by, sponsored by, or certified by Oracle Corporation.
