Qualixar OS: A Universal Operating System for AI Agent Orchestration

Varun Pratap Bhardwaj

cs.AIApr 7, 2026

Qualixar OS: A Universal Operating System for AI Agent Orchestration

Authors: Varun Pratap Bhardwaj

Organizations: Independent Researcher, Solution Architect, India

Abstract

We present Qualixar OS, the first application-layer operating system for universal AI agent orchestration. Unlike kernel-level approaches (AIOS) or single-framework tools (AutoGen, CrewAI), Qualixar OS provides a complete runtime for heterogeneous multi-agent systems spanning 10 LLM providers, 8+ agent frameworks, and 7 transports. We contribute: (1) execution semantics for 12 multi-agent topologies including grid, forest, mesh, and maker patterns; (2) Forge, an LLM-driven team design engine with historical strategy memory; (3) three-layer model routing combining Q-learning, five strategies, and Bayesian POMDP with dynamic multi-provider discovery; (4) a consensus-based judge pipeline with Goodhart detection, JSD drift monitoring, and alignment trilemma navigation; (5) four-layer content attribution with HMAC signing and steganographic watermarks; (6) universal compatibility via the Claw Bridge supporting MCP and A2A protocols with a 25-command Universal Command Protocol; (7) a 24-tab production dashboard with visual workflow builder and skill marketplace. Qualixar OS is validated by 2,821 test cases across 217 event types and 8 quality modules. On a custom 20-task evaluation suite, the system achieves 100% accuracy at a mean cost of $0.000039 per task. Source-available under the Elastic License 2.0.

Figures & tables

Panel of Figure 1 — Figure 1: Full component architecture of Qualixar OS . The core engine (center, yellow border) houses the Orchestrator’s 12-step pipeline with Forge, Swarm, Judge, Router, RL Trainer, and Cost Tracker. Seven transport channels (left) provide universal access. Quality guards (right, dashed green border) are new in Pivot 2 and emit events to the central EventBus/Dashboard monitoring stack (far right). Infrastructure spans the bottom. Red arrow: the reject $\to$ redesign feedback loop. Dashed arrows: event and feedback flows.

Panel of Figure 2 — Figure 2: End-to-end task lifecycle in Qualixar OS . Numbered steps 1–11 trace the 12-step pipeline from user input through transport, memory injection, Forge team design, model discovery and routing, swarm execution, and judge assessment. The diamond decision point routes to either RL learning and output (green path) or redesign (red loop, max 5 iterations). Quality monitors (blue sidebar) run in parallel during execution and feed guard verdicts into the judge assessment.

Feature	Companion	Power
Topologies	6	All 12
Max judges	2	5
Routing strategies	3	5 (+POMDP, balanced)
Reinforcement learning	No	Yes
Container isolation	No	Yes
Simulation	No	Yes

Table 1: Feature gates by operating mode.

Panel of Figure 3 — Figure 3: Model discovery and routing architecture. Configuration defines provider endpoints; the discovery engine queries 10 provider APIs at startup to build a live catalog of 236+ models with quality scores and pricing. Three routing strategies select models based on the task budget. Results cache with configurable TTL, and RL reward signals update strategy weights over time.

Provider	Discovery API	Auth
Azure AI Foundry	/models?api-version=...	Azure AD
OpenAI	/v1/models	API key
Anthropic	/v1/models	API key
Google (Vertex)	/v1/models	OAuth2
Bedrock	ListFoundationModels	AWS IAM
Ollama	/api/tags	Local

Table 2: Model discovery: supported providers and their catalog APIs.

#	Topology	Execution	Termination	Lines
1	Sequential	Chain: $A_{i}$ output $\rightarrow$ $A_{i+1}$ input	Last agent completes	35
2	Parallel	Fan-out via Promise.allSettled	All complete	40
3	Hierarchical	Manager decompose $\rightarrow$ workers $\rightarrow$ merge	Manager approves	60
4	DAG	Topological sort, level-parallel	All leaves complete	95
5	Mixture	$N{-}1$ generators $\rightarrow$ 1 aggregator	Aggregator completes	55
6	Debate	Proposer-critic rounds, “CONSENSUS” check	Consensus or max	70

Table 3: The 12 execution topologies with their execution semantics.

Panel of Figure 4 — Figure 4: Eight-module quality assurance pipeline. Each module emits typed events to the central EventBus. Rejected verdicts trigger the Forge redesign loop (left arrow). The Goodhart detector, drift monitor, and trilemma guard collectively prevent metric gaming and distributional shift.

Metric	Value
Source files (.ts + .tsx)	150+
Test cases (pass)	2,821
TSC errors	0
Database tables	49 (+1 FTS5, +1 meta)
API endpoints	60+
Dashboard tabs	24

Table 4: Qualixar OS system metrics (v2.0.0, April 2026).

Level	Tasks	Correct	Accuracy
Level 1 (factual, arithmetic)	7	7	100%
Level 2 (multi-step inference)	7	7	100%
Level 3 (probabilistic, complex)	6	6	100%
Overall	20	20	100%

Table 5: QOS Evaluation Suite: accuracy by difficulty level.

Metric	Value
Tasks	10
Iterations per task	3
Model	gpt-5.4-mini (Azure AI Foundry)
Mean final score	0.519
Tasks improved ( $\Delta>0$ )	3/10
Tasks converged (score $\geq 0.8$ )	6/10

Table 6: Loop benchmark: convergence analysis (10 tasks

\times

3 iterations, gpt-5.4-mini on Azure AI Foundry).

Panel of Figure 5 — Figure 5: Forge $\to$ Judge $\to$ RL loop convergence on a 10-task benchmark (gpt-5.4-mini). Shaded region indicates $\pm 1$ s.d. The downward trend is not statistically significant ( $p=0.578$ , paired $t$ -test); see Section 11.4 for interpretation.

Feature	AIOS	AutoGen	CrewAI	LangGraph	Qualixar OS
Topologies	N/A	2	2	DAG	12
Auto team design	No	No	No	No	Yes (Forge)
Cost routing	No	No	No	No	3-layer
Model discovery	No	No	No	No	10 providers
Quality judges	No	No	No	No	Consensus
Goodhart detection	No	No	No	No	Yes

Table 7: Feature comparison with related systems (v2.0.0, updated for Pivot 2).

Multi-Agent Orchestration

Multi-agent orchestration coordinates specialized AI agents, tools, and workflows to solve complex tasks efficiently and reliably.

See all 206 papers

The Agent Operating System (AOS): A Reference Operating Architecture for Distributed Agentic Systems
Aug 4, 2026Ankur Sharma, Deep ShahLLM Agent Orchestration AI Agent Governance
AOS Specification Series
Verified Multi-Agent Orchestration: A Plan-Execute-Verify-Replan Framework for Complex Query Resolution
Mar 12, 2026Xing Zhang, Yanwei Cui, Guanghui Wang +7Multi-Agent Orchestration LLM Agent Planning
AWS Generative AI Innovation Center · HSBC
PerspectiveGap: A Benchmark for Multi-Agent Orchestration Prompting
Jun 7, 2026Youran Sun, Xingyu Ren, Kejia Zhang +2Multi-Agent Orchestration LLM Agent Evaluation
University of Maryland · The Chinese University of Hong Kong · Stanford University

Aug 4, 2026J/K move · Enter open · S save

The Agent Operating System (AOS): A Reference Operating Architecture for Distributed Agentic Systems

Ankur Sharma, Deep Shah

AOS Specification Series

LLM Agent Orchestration AI Agent Governance LLM Agent Reliability

Open paper AskPDF arXiv

Large language models have transformed artificial intelligence from isolated prediction services into components of long-running, distributed systems that reason, invoke tools, retrieve external state, delegate tasks, and act on behalf of users and organizations. The surrounding ecosystem has responded with agent frameworks, workflow engines, model-serving platforms, memory systems, communication protocols, and observability tools. These technologies improve execution, but they do not provide a stable, implementation-independent operating architecture for governing intent, selecting capabilities, preserving authority across delegation, controlling uncertainty, coordinating runtime behavior, and reconstructing why consequential actions occurred. This paper proposes the Agent Operating System (AOS), a vendor-neutral reference operating architecture for distributed agentic systems. AOS contains two internal planes: a Control & Governance Plane responsible for intent, policy, trust, authority, confidence, auditability, observability, and human oversight; and a Runtime & Coordination Plane responsible for agent lifecycle, workflow coordination, model and tool routing, context and memory coordination, scheduling, traffic management, and runtime assurance. Platform services, Linux or Windows, container runtimes, and physical infrastructure remain outside the AOS boundary and are integrated through explicit interfaces. The paper specifies AOS concepts, invariants, interface objects, optimization objectives, deployment profiles, and reliability responsibilities. It also identifies tradeoffs and unresolved research questions. AOS is not presented as a replacement for existing frameworks or infrastructure; it is proposed as the operating architecture through which heterogeneous components can be composed into governable, reliable, observable, and interoperable agentic systems.

Abstract

Multi-Agent Orchestration

Explore similar work

The Agent Operating System (AOS): A Reference Operating Architecture for Distributed Agentic Systems

Verified Multi-Agent Orchestration: A Plan-Execute-Verify-Replan Framework for Complex Query Resolution

PerspectiveGap: A Benchmark for Multi-Agent Orchestration Prompting