Share what's on your mind
Your decisions are held

$_

Ops-grade AI, not just “answers”. Reproducible runs. Auditable evidence. Safe execution.

MARIA CODE (personal) / PLANET (enterprise)
QUALITY.BY.DEFAULT
PREVENT INCIDENTS, NOT JUST ERRORS
No silent regressions
Safety gates + fail-closed defaults when evidence is missing.
Reproducible by design
Artifacts + integrity refs; long-term memory is evidence-backed.
Anti-drift architecture
Multi-layer SSOT makes drift visible, enforceable, and recoverable.
ENTERPRISE.ASSESSMENTPLANET Readiness

Diagnose readiness to deploy PLANET in your organization.
120 yes/no questions. Instant KFS1–KFS6 score, bottleneck, and a 90-day roadmap.

RUN.ASSESS
KFS1 MODELKFS2 AGENTSKFS3 DATAKFS4 OPSKFS5 EXECKFS6 EVOLVE
EVIDENCE.FIRST
REPRODUCIBLE.RUNS
SAFE.DEFAULTS
WHAT.IS.UNIVERSE?

Universe is the persistent container for Maria OS.

A Universe holds operating history as SSOT: runs, artifacts, decisions, evidence, and the evolving structure behind them. It is designed for reproducibility and safe iteration (SIM-first, policy-aware, audit-ready).

UNIVERSE.OS creates and manages that container.
Universe Visual Playground lets you inspect execution (planets/envelopes) and data graphs in 3D, without inventing facts (UI reads SSOT views / refs only).
Note: The Playground requires sign-in (Google / GitHub).
SSOT.SHAPE
Runs → Artifacts → Evidence
Every run produces immutable refs. The UI renders them; it does not compute truth.
3D Execution Layer
Planets (agents/commands) and envelopes (A2A) are visualized with LOD and performance budgets.
Data Universe Layer
Nodes/edges are read from SSOT SQL views (vw_unv_*) and overlaid as a separate layer.

COMMAND MATRIX

Commands that prevent incidents: evidence, gates, and reproducible outcomes.

1. GOVERNANCE (Decision & Structure)
PLANET
$ /structure "We keep breaking CI after dependency upgrades"

Prevents recurring incidents by turning “it keeps happening” into explicit policies, gates, and owner-bound actions.

OUTPUT: Outputs DecisionTrace + policy proposal + evidence refs (rollback criteria included).
PLANET
$ /cxo "Should we ship feature X this week?"

Prevents risky “ship-now” decisions without a record of tradeoffs, owners, and stop conditions.

OUTPUT: Outputs DecisionTrace (go/no-go) + risk/mitigations + evidence refs (auditable later).
2. AGENTS (Organization & Execution)
PLANET
$ /agents init "Improve ATS performance" --product ats

Prevents coordination chaos by creating explicit roles, envelopes, and execution boundaries before work starts.

OUTPUT: Creates an agent team (roles + envelopes + scope) and a traceable execution plan with evidence refs.
PLANET
$ /agent auto --source cxo --last 20

Prevents “lost intent” by executing from decisions with preserved context, constraints, and artifact outputs.

OUTPUT: Executes with coordination and produces artifacts + EvidenceIndex refs for review.
3. EXECUTION (Maria OS & Build)
PLANET
$ /universe init --org org_123 --name default

Prevents institutional memory loss by creating a persistent container for runs, evidence, and learning.

OUTPUT: Creates a Universe where runs, artifacts, evidence, and checkpoints can be preserved and replayed.
ALL PLANS
$ /code "Fix the failing tests and keep behavior unchanged"

Prevents “quick fixes” that silently change behavior by making intent explicit and verifying results.

OUTPUT: Generates a reviewable diff + verification steps; produces GateReport when gates are run.
PLANET
$ /auto-dev run "Refactor utils for readability"

Prevents unreviewed autonomous changes by enforcing safety gates, evidence, and controlled scope.

OUTPUT: Applies changes, runs required gates, emits GateReport, and preserves EvidenceIndex + artifacts for audit.
4. EVOLUTION (Learning Loop)
PLANET
$ /evolve "Close P0 issues from doctor findings" --dry-run

Prevents quality decay by running a repeatable diagnose → decide → execute → verify → learn loop.

OUTPUT: Produces TaskSpecs + GateReport + EvidenceIndex; checkpoints allow partial replay of long runs.
5. HEALTH (Safety & Diagnostics)
PLANET
$ /doctor

Prevents late discovery by diagnosing issues early with evidence and a structured risk view.

OUTPUT: Outputs findings with evidence refs and risk tiers (reviewable, reproducible diagnostics).
6. MEDIA (Image & Video Generation)
ALL PLANS
$ /image "Modern admin dashboard UI"

AI-powered image generation

OUTPUT: Generates professional graphics with intelligent routing
ALL PLANS
$ /video "30-second product demo"

AI video generation

OUTPUT: Creates product demos and presentations
MARIA.TERMINAL.v3.6.0
$ maria /evolve "Close P0 issues from doctor findings" --dry-run
[DIAGNOSING] Running doctor scan...
[ANALYZING] Found 3 P0 issues with evidence
[DECIDING] Classifying safe/guarded/risky...
[PLANNING] Generating TaskSpec with constraints...
[EXECUTING] Agents coordinating: implementation → testing → review
[VERIFYING] GateReport: schema validation ✓ CI passed ✓ tests increased
[LEARNING] DoctorDelta: improvement confirmed, updating Maria OS...
✓ EVOLVE COMPLETE: 3 P0 issues resolved
✓ EVIDENCE: DecisionTrace + GateReport generated
✓ MEMORY: Maria OS updated with lessons learned
✓ READY: System improved, reproducible re-runs available

SYSTEM CAPABILITIES

PLANET: capabilities designed to prevent drift, regressions, and un-auditable work.

PLANET ENTERPRISE FEATURES
PLANET

GOVERNANCE.LAYER

Prevents unreviewable decisions by enforcing explicit criteria, boundaries, approvals, and deterministic gates.

/structure, /cxo
PLANET

AGENTS.ORCHESTRATION

Prevents coordination failure by making roles, scope, and handoffs explicit (contract-driven execution).

/agents, /agent
PLANET

UNIVERSE.MEMORY

Prevents context loss and irreproducible work by preserving runs, artifacts, and evidence as SSOT (Data.OS bound by default).

/universe
PLANET

GIT.INTELLIGENCE

Prevents unsafe Git operations by enforcing read-only inspection and safety constraints, then routing work with evidence.

/git
PLANET

GIT.CULTURE.LAYER

Prevents “work only on a laptop” by treating Git as a ledger: EvidenceIndex + push-pending + pseudo PR when publish is blocked.

/git-culture
PLANET

DATA.SCIENCE.WORKFLOWS

Prevents unverifiable analysis by keeping SQL-first understanding structured (chunk → merge) with explicit assumptions and evidence.

/ds
PLANET

EVOLUTION.LOOP

Prevents long-term quality decay by institutionalizing diagnose → decide → execute → verify → learn.

/evolve
PLANET

HEALTH.DIAGNOSTICS

Prevents late discovery by diagnosing issues early with evidence, risk tiers, and actionable outputs.

/doctor
PLANET

SECURITY.COMPLIANCE

Prevents cross-tenant leakage with isolation, audit logs, and compliance-ready architecture.

SOC2/HIPAA
ADDITIONAL CAPABILITIES (ALL PLANS)
ALL PLANS

CODE.GENERATION

Prevents slow context switching by turning intent into a reviewable patch quickly.

/code command
ALL PLANS

INTELLIGENT.ROUTING

Prevents fragile “model ops” by routing capabilities automatically without exposing model names.

Auto routing
ALL PLANS

IMAGE.GENERATION

Prevents design bottlenecks for basic assets (fast drafts for teams).

/image command
ALL PLANS

VIDEO.CREATION

Prevents demo production becoming a blocker (generate draft demos fast).

/video command

Maria OS

Two products, one philosophy.
MARIA CODE for individuals: a developer AI agent. Maria OS for enterprises: the place where leadership is held.
Maria OS exists for leaders who carry responsibility in silence. It holds your decisions, preserves your way of thinking, and provides presence without intrusion.

PERSONAL

MARIA.CODE (DEV.AGENT)

Move faster in building, debugging, learning, and shipping with command-level execution.

  • • Fast code generation & refactors
  • • Local + cloud model routing
  • • Documentation → execution (/help)
ENTERPRISE

Maria OS (ORG.OPERATIONS.OS)

PLANET USER ONLY

The place where your decisions are held, preserved without judgment. Your way of thinking, your priorities, your long-term instincts—they remain here, even when you are tired, even when you are gone.

  • • Governance, roles, and approval loops (who can do what, when, and why)
  • • Data boundaries, access control, and auditability
  • • Local-first execution when required (enterprise aligned with LOCAL_MODE)
  • • Deterministic outputs with evidence: DecisionTrace + GateReport + reproducible re-runs
ENTERPRISE.FEATURES
  • Connectors: Seamless integration with GitHub, accounting/ERP, Google, and more. CEO-managed connector settings for secure data synchronization.
  • Admin Management: Tenant-scoped member management with role-based access control (CEO, CFO, COO, CAIO, MEMBER, GUEST).
  • Tenant & Role Management: Multi-tenant architecture with granular permissions. Invite members, assign roles, and control data access boundaries.
  • Maria OS Seamless Access: Unified memory system with event tracking, lessons learned, and reproducible procedures. Cross-tenant isolation with secure access patterns.
  • Security & Compliance: Military-grade encryption, audit logs, and compliance-ready architecture. No cross-tenant data leakage. SOC2/HIPAA aligned.
WHY.MARIA.OS (WHY NOW)
THE OS LAYER WINS
Models are commoditized
Everyone gets access. Advantage shifts from model choice to how reliably your org can operate with AI.
Risk is the bottleneck
Without structure, you get speed without control. Maria OS makes safety and traceability the default.
Adoption compounds
Tools don’t compound. Operating loops do. The earlier you install the OS layer, the faster it compounds.
Maria OS exists for leaders who carry responsibility in silence. When “AI” stops being a feature and becomes a production surface, the question is no longer “Can we generate?” but “Can we hold decisions, preserve context, and continue improving?” The OS layer holds your decisions: clear boundaries, preserved context, and loops that keep quality rising while risk falls. Your decisions are held here, preserved without judgment. When you return days or months later, they are still there. Unchanged. Unjudged. Starting with MARIA OS early allows you to preserve your way of thinking and priorities from the beginning, so your organization can grow without burning out.
FAILURE.MODES → PREVENTION.MECHANISMS
QUALITY IS A SYSTEM, NOT A WISH
Silent regressions shipped
Required gates + independent judgment (Auto-Dev executes, QE decides) + fail-closed defaults.
Cannot reproduce runs
Checkpoints + partial replay + artifact integrity (hash-verified refs, cache safety).
Work exists only on a laptop
Git as Ledger + EvidenceIndex; when push/PR is blocked, emit Push Pending + pseudo-PR for recovery.
Fuzzy logic in core enforcement
No-heuristics boundary (D0–D3): probabilistic inference must use an LLM; if unavailable, stop with reasons.
Local execution is flaky
Local provider manager (autostart + diagnostics + compatibility retries) so ops does not depend on hero machines.
Quality decays over time (drift)
Multi-layer SSOT (docs/config/src/schemas/artifacts) to make drift visible, enforceable, and recoverable.
Maria OS in action (current)
Example prototype: GitHub Code Review Maria OS — comment-only PR reviews with evidence, deterministic rules, and idempotent posting, plus DeliveryOps-ready deliverables generation. Your decisions and their outcomes are preserved here.
You can start small (one team, one workflow) and preserve patterns as your organization grows. Your decisions are held, not judged.
ENTERPRISE.FEATURES.DETAIL

Enterprise-Grade Capabilities for Organizational Intelligence

CONNECTORS

Seamlessly integrate with external systems to populate your CXO dashboard and sync organizational data.

  • • Integration with external systems (GitHub, accounting/ERP, Google, and more)
  • • CEO-managed connector settings
  • • Secure OAuth2 authentication
  • • Real-time synchronization
  • • Audit logs for all connector activities
ADMIN MANAGEMENT

Comprehensive tenant-scoped member management with role-based access control and granular permissions.

  • • Tenant-scoped member management
  • • Roles: CEO, CFO, COO, CAIO, MEMBER, GUEST
  • • Custom entitlements
  • • Data scope boundary configuration
  • • Invitation link generation with expiry management
TENANT & ROLE MANAGEMENT

Multi-tenant architecture with strict isolation. Each tenant operates independently with its own members and data boundaries.

  • • Multi-tenant architecture
  • • Tenant isolation (no cross-tenant data reads)
  • • Automatic CEO role assignment
  • • Member invitation and onboarding
  • • Role-based access control
UNIVERSE SEAMLESS ACCESS

Unified memory system with event tracking, lessons learned, and reproducible procedures. Cross-tenant isolation with secure access patterns.

  • • Unified memory system (Event → Lesson → Procedure)
  • • DecisionTrace + GateReport
  • • Reproducible re-runs
  • • Automated event tracking and learning
  • • Cross-tenant isolation with secure access patterns
SECURITY & COMPLIANCE

Military-grade encryption, comprehensive audit logs, and compliance-ready architecture. SOC2/HIPAA aligned.

  • • End-to-end encryption
  • • Comprehensive audit logs (ACCESS.CONTROL / AUDIT.LOGS panels)
  • • No cross-tenant data leakage
  • • LOCAL_MODE support (local-first execution option)
  • • SOC2/HIPAA compliant architecture
GOVERNANCE.LAYER
Policy, permissions, approvals, and compliance-ready logs—built into the workflow by default.
OPS.LOOPS
Reproducible runs, runbooks, and safety rails that prevent silent failures and “it worked yesterday” drift.
ORG.ADOPTION
Deploy “roles” (not tools). Standardize patterns across teams via shared contracts.
SPEED × QUALITY
Increase decision speed while reducing risk via explicit structure and traceability.

ROUTING NEURAL NETWORK

Profiles & capabilities are selected automatically. No model names exposed.

FAST

[ONLINE]

Latency: low

BALANCED

[ONLINE]

Latency: medium

QUALITY

[ONLINE]

Latency: medium

VISION

[ONLINE]

Latency: medium

IMAGE

[ONLINE]

Latency: medium

VIDEO

[ONLINE]

Latency: medium

LOCAL

[OPTIONAL]

Latency: depends

FALLBACK

[ACTIVE]

Latency: auto

INSTALLATION

Cross-platform support for Windows, macOS, and Linux

QUICK.START

Global Installation
$ npm install -g @bonginkan/maria
Start MARIA
$ maria
Check Version
$ maria --version

ADVANCED.OPERATIONS

Update to Latest
$ npm update -g @bonginkan/maria
Force Reinstall
$ npm install -g @bonginkan/maria --force
Uninstall
$ npm uninstall -g @bonginkan/maria

PLATFORM.SUPPORT

🪟 Windows✓ FULLY SUPPORTED
🍎 macOS✓ FULLY SUPPORTED
🐧 Linux✓ FULLY SUPPORTED
Node.js Requirement≥ 20.10.0

TROUBLESHOOTING

EACCES Permission Error (Common with Homebrew)?

sudo chown -R $(whoami):admin /opt/homebrew/lib/node_modules/@bonginkan/maria
Then run: npm install -g @bonginkan/maria@latest

Alternative: Use sudo (Quick Fix)

sudo npm install -g @bonginkan/maria

Fix npm permissions (recommended):

npm config set prefix ~/.npm-global

Add to PATH (Linux/Mac):

export PATH=~/.npm-global/bin:$PATH

✅ After Installation

🏠 From Home Directory:

cd ~/ && maria /help

📁 From Any Project:

cd /path/to/project && maria /code create component

💻 VS Code Terminal:

maria --version

🔄 Check Alias:

which maria

⚠️ If you see "aliased", you might be using a local development version instead of global

Need help? Check the documentation or contact us via contact

LOCAL LLM INFRASTRUCTURE

Run powerful models on your own hardware - No cloud dependency

OLLAMA

  • • Llama 3.2 (3B, 70B)
  • • Mistral (7B, 8x7B)
  • • DeepSeek Coder
  • • Phi-3
  • • Auto model management

LM STUDIO

  • • Large open-source model
  • • Lightweight open-source model
  • • GUI model management
  • • GPU acceleration
  • • OpenAI-compatible API

vLLM

  • • High-throughput serving
  • • Tensor parallelism
  • • Continuous batching
  • • PagedAttention
  • • Production-ready

Full compatibility with local inference servers • Zero-config integration • Automatic fallback

Local LLM Setup Guide (Ultra–Enterprise)

PRICING MATRIX

Choose your development acceleration tier

Ops-grade reliability is included.
Every tier is built on the same quality philosophy: evidence-first workflows, reproducible runs, and safe defaults.

🚀 ALL PLANS NOW AVAILABLE!

All plans including Starter, Pro, and Ultra are now available for immediate activation.
Sign in with Google to get started, then upgrade to your preferred plan.

Free

Free plan for trials and learning

Free
Basic AI Chat
Code Generation (40/month)
5 file analysis/month
Intelligent routing (Basic)
Community Support
Popular
LIVE NOW

Starter

Standard plan for individual developers

$20/month
Advanced AI Chat
Code Generation (300/month)
50 file analysis/month
Intelligent routing (Standard)
Community support
LIVE NOW

Pro

For individuals & small teams (recommended)

$39/month
Professional AI Chat
Advanced Code Generation (1,200/month)
200 file analysis/month
Intelligent routing (Advanced)
Community support
API Access
LIVE NOW

Ultra

For power users (research / production)

$99/month
Executive AI Chat
Professional Code Generation (5,000/month)
500 file analysis/month
Intelligent routing (Premium)
Priority support
API Access

All plans include web dashboard, API access, and usage analytics

CODE AT THE SPEED OF THOUGHT

Transform ideas into production-ready code instantly.

NO.CREDIT.CARD // INSTANT.ACCESS // FREE.TIER.AVAILABLE