Captain and Crew: Agent-to-Agent Collaboration in ketteQ’s Quintus™

Supply chain questions rarely stay in one lane. A shift in demand changes the inventory position, which changes the supply plan, which runs into capacity. Answering well means working across all of those functions at once, at the pace the business requires.

Quintus™, ketteQ’s Free-Range AI™ for supply chain, does this through agent-to-agent collaboration. Instead of asking one generalist model to cover every domain, Quintus coordinates a crew of specialized agents, each focused on one part of the planning process, and brings their work together into a single answer or action. The result is faster, better planning from agents that learn their jobs and adapt as the organization changes.

Collaboration in Quintus runs in two directions: internally, across ketteQ’s specialist agents, and externally, with the agents customers have already built or bought.

The Captain and Crew Model

Quintus acts as the captain. It takes in the request, determines which parts of the business it touches, assigns the work to the right specialists, and reconciles what comes back. The crew is a set of fit-for-purpose agents for forecasting, demand planning, inventory, supply planning, capacity, and other functions, each built around how that function runs in your business.

Take a demand question. If it is simple, Quintus answers it directly. If it affects several functions, as most planning questions do, Quintus routes it through the crew. The forecasting agent evaluates several statistical methods and returns the best fit. The demand planning agent assesses whether the change represents a meaningful lift. The supply and inventory agents test the result against the network. If the new forecast would exceed capacity next quarter, that conflict goes back to Quintus, which brings in the capacity agent to determine what is feasible. The planner receives one reconciled answer rather than four separate analyses to stitch together.

Authority stays with the captain. Specialists do not vote or negotiate with one another. Quintus directs the work and makes the call, within the guardrails the organization has defined.

Why Specialization Is More Efficient

A specialist works with the data, methods, and business rules of its own domain, so it reaches a reliable answer faster than a generalist covering everything. The captain absorbs the coordination work that slows planning teams down: handoffs between functions, reconciliation cycles, and rework when one team’s plan breaks another’s.

The model also scales cleanly. Expanding what Quintus can do means adding or extending a specialist, not retraining a single model to do more.

Agents That Grow With Your Organization

The crew is not static. Each specialist learns its task and continues to evolve as the business does: when policies and priorities shift, when new products, channels, sites, or suppliers come online, and when the way your teams plan changes. Over time the crew reflects how your organization operates today, not how it operated on the day the system went live.

That matters most when conditions change quickly. When a disruption hits, whether a supplier failure, a demand spike, or a logistics constraint, Quintus mobilizes the specialists the situation requires, and the crew can be extended as new kinds of problems emerge. Instead of fixed workflows that must be reconfigured every time the business changes, organizations get a team of agents that grows with them.

Working With the Agents You Already Have

Most enterprises are already investing in agents of their own, whether inside ERP, CRM, procurement, and logistics platforms or built in-house on their own AI stack. Quintus is designed to put those investments to work, not to replace them.

Quintus uses two open protocols here, and they solve different problems. The Model Context Protocol (MCP) gives an agent access to systems: it can query data and call defined functions, but the reasoning stays with the agent making the request. Agent2Agent (A2A), originally developed by Google and now governed by the Linux Foundation, connects one agent to another. When Quintus engages a customer’s agent over A2A, it is not reading that agent’s data. It is asking for that agent’s judgment, applied with its own models, context, and business rules. MCP extends Quintus’s reach into a customer’s systems. A2A extends its reach into the expertise a customer has already invested in.

In practice, that means a customer’s agents can contribute to the plan the way Quintus’s own specialists do. A procurement agent can weigh in on which suppliers are at risk and which alternates to qualify. A sales agent can offer its view of which pipeline will actually convert. Quintus factors that judgment into its decisions, and when a plan changes, it can work through the agents that already own downstream execution rather than requiring new integrations or manual handoffs. The relationship also works in reverse: a customer’s agent can hand Quintus a planning task and get a reasoned answer back.

Because A2A is an open standard, none of this requires ketteQ to build a custom bridge for each customer’s environment. Every agent a customer has already deployed becomes a potential member of the crew.

Governed by Design

Quintus is unscripted and unconstrained in how it reasons, and fully governed in how it acts. Every Quintus agent operates inside guardrails set at configuration time, and the system actively monitors for conflicting or duplicate actions across the agents it coordinates, flagging or resolving them before they become a problem.

PolymatiQ™, ketteQ’s patent-pending agentic solver, is not one of the agents in the conversation. It is the engine the agents call to run scenarios, so the underlying math stays consistent instead of being rebuilt for every request.

Built on the Systems You Already Run

None of this requires a rip-and-replace. Quintus connects to a customer’s ERP and planning systems, SAP or otherwise, through connectors built on dlt, an open-source, Python-native data pipeline tool that extracts and normalizes data and can write results back. Existing systems remain the source of data; Quintus provides the reasoning and coordination on top.

Why Architecture Matters

Multi-agent AI is now a common theme across supply chain software, but much of what has been announced is still on roadmaps. Coordinating agents in production, and connecting to agents outside a vendor’s own platform, is harder than launching an assistant. Architecture also determines reach: a platform built on open protocols such as A2A can work across a customer’s broader agent ecosystem, while a closed system can only coordinate with itself.

ketteQ’s platform was built on an open stack of Python, SQL, and JSON, not the rigid architectures many legacy planning systems are now retrofitting with AI. That foundation let the team stand up a crew of specialist agents quickly, and it is what lets that crew keep expanding.

Seeing It in Production

Alliance Consumer Group (ACG) is already putting this model to work with Quintus™. By bringing intelligence across demand, supply, inventory, and execution together, Quintus helps ACG move from questions to decisions faster without requiring teams to manually reconcile information across systems. More than 200 field sales representatives can also request Available-to-Promise directly within Salesforce and receive an accurate promise date in seconds.

“We were spending too much time reconciling spreadsheets and answering one-off questions,” said Stephanie Larson, Director of Supply Chain at ACG. “What we needed was real-time visibility that everyone could trust and act on.”

With Quintus coordinating intelligence behind the scenes, ACG can spend less time stitching together answers and more time acting on them

What This Means for Planning Teams

Agent-to-agent collaboration in Quintus organizes AI the way strong operating teams already work: a clear point of command, specialists who know their jobs and get better at them, and the ability to bring in outside expertise when the situation calls for it. The result is planning that moves faster, adapts as the business changes, and makes full use of the AI investments an organization has already made.