Every question, answered.

What is Trunnion AI?

Trunnion AI is an enterprise-grade, LLM-agnostic multi-agent orchestration platform built by Viceroy NM. It coordinates teams of AI agents across any large language model, any classification level, and any deployment environment — including air-gapped infrastructure — with governance, audit trails, and human-in-the-loop controls built into the architecture.

Who built Trunnion AI?

Trunnion AI was built by Viceroy NM, a New Mexico-based technology company specializing in governed agentic AI infrastructure for enterprise and defense customers. Viceroy NM also develops vertical SaaS products powered by the Trunnion AI framework.

When was Trunnion AI founded?

Viceroy NM, the parent company of Trunnion AI, was founded in 2024. The Trunnion AI orchestration framework has been in active development since then and powers a growing portfolio of vertical SaaS products in production.

What does Trunnion AI do?

Trunnion AI orchestrates teams of AI agents to plan and execute multi-step workflows autonomously. It handles agent coordination, tool authorization, classification enforcement, human-in-the-loop approvals, and full audit logging for regulated and mission-critical environments.

What problem does Trunnion AI solve?

Trunnion AI solves the problem of deploying autonomous AI agent teams in environments where governance, classification controls, and human oversight are non-negotiable — like defense, government contracting, regulated manufacturing, and enterprise infrastructure. Most AI platforms treat governance as an afterthought; Trunnion AI builds it in as architecture.

Where is Trunnion AI based?

Trunnion AI is headquartered in Albuquerque, New Mexico, USA. It is built and owned by Viceroy NM, LLC.

Is Trunnion AI a chatbot?

No. Trunnion AI is enterprise infrastructure for deploying autonomous AI agents that execute multi-step workflows end-to-end. Chatbots respond to prompts; Trunnion AI plans goals, selects tools, calls APIs, completes long-running tasks, and operates with audit trails and policy enforcement.

Is Trunnion AI an LLM?

No. Trunnion AI is an orchestration platform that runs on top of LLMs. It coordinates AI agents that use LLMs (OpenAI GPT, Anthropic Claude, Google Gemini, Mistral, Ollama-hosted models) to perform reasoning, but it is not itself a language model.

What does the name Trunnion mean?

A trunnion is a structural pivot — historically the cylindrical pin on a cannon that allowed it to rotate while remaining anchored to its mount. The name reflects the platform's role as a stable structural foundation that gives AI agents the ability to rotate and adapt across tasks while remaining anchored to governance and policy.

Is Trunnion AI a startup?

Yes. Trunnion AI is a privately-held venture from Viceroy NM. The company operates with a vertical-AI strategy — using its core orchestration framework to power multiple production SaaS products across defense, GovCon, manufacturing, real estate, events, and IoT logistics.

How does Trunnion AI work?

Trunnion AI uses a 3-tier agent architecture. An orchestrator agent receives a task, decomposes it into sub-tasks, and assigns each to specialist agents. Specialists invoke executor agents to perform discrete actions like API calls or database queries. Every action passes through policy validation, classification enforcement, and audit logging.

What is multi-agent orchestration?

Multi-agent orchestration is the coordination of multiple AI agents working together on a complex task. Each agent has specialized capabilities — research, analysis, validation, execution. An orchestrator agent decomposes the task, assigns sub-tasks, manages dependencies, handles failures, and synthesizes results into a final output.

What is the 3-tier agent architecture in Trunnion AI?

Tier 1 is the orchestrator — receives tasks, plans the workflow, manages state. Tier 2 is specialists — domain experts (procurement specialist, compliance specialist, etc.) that handle sub-tasks. Tier 3 is executors — narrow-scope agents that perform single actions like calling SAM.gov, validating CAC credentials, or writing audit log entries.

What does LLM-agnostic mean?

LLM-agnostic means Trunnion AI is not locked into any single AI model provider. It orchestrates agents running on OpenAI, Anthropic, Google Gemini, Mistral, or locally-hosted open-source models via Ollama. Organizations can switch or mix models without re-architecting their workflows.

Why is LLM-agnosticism important?

LLM-agnosticism prevents vendor lock-in, lets organizations route different tasks to the best model (e.g., reasoning to Claude, coding to GPT, classified work to local Ollama), enables compliance with sovereignty requirements, and protects against pricing or availability changes from any single provider.

What LLMs does Trunnion AI support?

Trunnion AI supports OpenAI (GPT-4 and successors), Anthropic Claude (Sonnet, Opus, Haiku), Google Gemini, Mistral, and any open-source model deployable via Ollama (Llama, Qwen, Mixtral, DeepSeek, etc.). Adding new providers requires no architectural changes.

Can Trunnion AI mix multiple LLMs in a single workflow?

Yes. A single workflow can route different agent roles to different LLMs based on cost, capability, latency, or compliance requirements. For example, a complex reasoning agent might use Claude Opus while a simple data extraction agent uses GPT-4o-mini.

What is policy-aware execution?

Policy-aware execution means agents enforce attribute-based access control, classification rules, and authorization policies on every action they take. Tool calls, data access, and outputs are validated against policy before execution. Violations are blocked and logged.

What is human-in-the-loop AI?

Human-in-the-loop (HITL) AI means a human operator reviews and approves certain agent actions before they execute. Trunnion AI builds HITL approval gates into every workflow at configurable risk thresholds. Routine tasks run autonomously; high-consequence actions require human approval.

How does Trunnion AI handle long-running workflows?

Trunnion AI uses durable state persistence backed by PostgreSQL and Redis-based task queues (BullMQ). Workflows can run for hours, days, or longer with full state recovery on failure. Each step is checkpointed, and execution can resume from any checkpoint.

Does Trunnion AI support parallel agent execution?

Yes. Workflows support parallel branching where multiple agents execute concurrently, with synchronization at join points. The orchestrator manages dependencies, fan-out/fan-in patterns, and conditional branching.

Is Trunnion AI NIST 800-53 compliant?

Yes. The DAF product includes NIST 800-53 compliance mapping with CAC/PIV identity traceability, five-level classification enforcement, and SHA-512 tamper-evident audit trails. Compliance is architectural, not bolt-on.

Does Trunnion AI support CMMC compliance?

Yes. Trunnion AI is architected for CMMC Level 2 and Level 3 readiness, with controls mapped to the underlying NIST SP 800-171 and 800-172 requirements. CMMC certification depends on the customer's overall environment, not just Trunnion AI alone.

What is attribute-based access control (ABAC) in Trunnion AI?

ABAC means access decisions are based on attributes of the user (clearance level, role, location), the resource (classification, sensitivity, owner), and the context (time of day, network, request type). Trunnion AI evaluates ABAC policies before every tool call or data access.

What is the six-layer tool authorization in Trunnion AI?

Trunnion AI's authorization pipeline checks: (1) user identity, (2) user clearance, (3) classification of input data, (4) classification of output data, (5) tool-specific policy, and (6) workflow-level policy. All six layers must pass before a tool executes.

What is a tamper-evident audit trail?

A tamper-evident audit trail uses cryptographic hash chaining (SHA-512 in Trunnion AI) so that any modification to a past log entry breaks the chain and is immediately detectable. Each entry contains the hash of the previous entry, creating an immutable sequence.

Can I see what an AI agent did and why?

Yes. Trunnion AI provides full reasoning traces, tool call logs, and execution replay. Every agent decision, tool invocation, input, output, and policy check is recorded. Analysts can replay any past execution step-by-step.

Does Trunnion AI support classification levels?

Yes. Five classification levels are supported in DAF: UNCLASSIFIED, CONFIDENTIAL, SECRET, TOP SECRET, TS/SCI. Classification is enforced on data, tool calls, and agent outputs. Cross-classification leakage is architecturally blocked.

How does Trunnion AI prevent classified data from leaking to unclassified systems?

Trunnion AI enforces classification at the data level, the agent level, the tool level, and the network level. Classified agents cannot invoke unclassified tools. Classified data cannot flow to unclassified outputs. Air-gap deployments physically separate classified workloads from external networks.

Does Trunnion AI support FedRAMP?

Trunnion AI's architecture is designed to be FedRAMP-deployable. Specific FedRAMP authorization status depends on the customer's deployment environment. Contact Viceroy NM for current FedRAMP roadmap and deployment options.

Is Trunnion AI ITAR compliant?

Trunnion AI's air-gap deployment mode and U.S.-only personnel access controls support ITAR compliance for technical data handling. Customers operating in ITAR-regulated environments can deploy Trunnion AI on-premises or in U.S.-only sovereign cloud regions.

Who uses Trunnion AI?

Trunnion AI is used by defense and intelligence agencies, government contractors (primes and subs), manufacturers, waste management operators, real estate brokerages, creative event vendors, beverage distributors, and other enterprise customers across the seven verticals Trunnion-powered products serve.

What industries does Trunnion AI serve?

Defense and intelligence, government contracting, manufacturing, smart waste and fleet operations, real estate, the creative events industry, and IoT logistics. Each vertical has a purpose-built product running on the Trunnion AI orchestration core.

Is Trunnion AI for small businesses or enterprises?

Trunnion AI's vertical SaaS products serve organizations of all sizes. RealtrAI works for individual real estate agents. Haultro scales from 25 to 50,000+ bins. ZennVue serves single photographers and large multi-vendor venues. The core Trunnion AI framework is enterprise-grade infrastructure for organizations needing governance and compliance.

What are common use cases for Trunnion AI?

Common use cases include autonomous procurement workflows, defense intelligence analysis, manufacturing production intelligence, IoT-driven fleet optimization, two-sided marketplace operations, AI-assisted real estate workflows, regulatory compliance automation, and any multi-step business process that benefits from coordinated AI agents.

Can I integrate Trunnion AI with my existing systems?

Yes. Trunnion AI integrates via REST APIs, GraphQL, webhooks, message queues, and direct database connectors. Pre-built integrations exist for SAM.gov, Google Workspace, Microsoft 365, Stripe, Salesforce, and major enterprise systems. Custom integrations are part of standard deployment.

Does Trunnion AI replace my employees?

No. Trunnion AI augments human work by handling repetitive, multi-step processes that consume hours of skilled labor. The HITL design ensures humans remain in command of high-consequence decisions. Customers typically report that Trunnion AI lets their teams focus on higher-value work.

What ROI do customers typically see from Trunnion AI?

Customers typically see 40–60% time reduction on automated workflows, 5–10x cost reduction versus incumbent platforms, and rapid time-to-value (production deployment within 90 days). Specific ROI varies by use case and deployment model.

Is Trunnion AI in production use today?

Yes. Multiple Trunnion-powered products are live in production: DAF, VPW, Cortrova, Haultro, ZennVue, and RealtrAI. FizzSense is in pre-launch with early access available.

What is DAF?

DAF (Declarative Agentic Framework) is Trunnion AI's defense product. It provides 150+ pre-built AI agents across 11 mission workflows, fully air-gap deployable with NIST 800-53 compliance mapping, CAC/PIV identity traceability, five-level classification enforcement, and SHA-512 hash-chain audit trails.

Who is DAF for?

DAF is built for the defense and intelligence community: DoD agencies, intelligence community organizations, defense primes (Lockheed, Raytheon, Northrop, Boeing, etc.), federally funded research and development centers (FFRDCs), and cleared defense contractors operating at SECRET, TOP SECRET, and TS/SCI levels.

How many agents does DAF include?

DAF includes 150+ pre-built AI agents organized across 11 mission workflows including intelligence analysis, threat assessment, mission planning, logistics optimization, supply chain analysis, and operational reporting.

Can DAF run air-gapped?

Yes. DAF supports fully air-gapped deployment using locally-hosted open-source LLM weights via Ollama. No external network connectivity is required. Suitable for SCIFs and classified network environments up to TS/SCI.

What classification levels does DAF support?

DAF supports five classification levels: UNCLASSIFIED, CONFIDENTIAL, SECRET, TOP SECRET, and TS/SCI. Classification enforcement is architectural and applies to data, agents, tools, and outputs.

Does DAF support CAC and PIV authentication?

Yes. DAF integrates with CAC (Common Access Card) and PIV (Personal Identity Verification) authentication for DoD and federal civilian environments. Identity attributes from CAC/PIV cards drive ABAC policy decisions throughout the platform.

What is the DAF visual workflow designer?

DAF includes a drag-and-drop visual workflow designer that lets analysts compose multi-agent workflows without writing code. Users can wire agents together, configure HITL gates, set classification controls, and deploy workflows directly from the UI.

How does DAF compare to Palantir Foundry for defense?

Palantir Foundry is a data integration platform with AI features added on top. DAF is a purpose-built multi-agent orchestration platform. DAF runs at 5–10x lower cost, deploys in weeks rather than months, supports any LLM (not just Palantir's choices), and provides pre-built agents for common defense mission workflows.

How long does DAF take to deploy?

A typical DAF deployment reaches production in 60–90 days, including environment setup, agent configuration, security accreditation support, and operator training. Air-gapped deployments may take longer based on facility access and accreditation timelines.

What time savings does DAF provide?

DAF customers typically report 40–60% time reduction on multi-step analytical and operational workflows compared to manual processes or legacy tooling, with full audit traceability maintained throughout.

Where can I learn more about DAF?

Visit trunnion.ai/daf for the complete DAF product overview, technical specifications, deployment options, and demo request.

What is VPW?

VPW (Viceroy Productivity Workstation) is Trunnion AI's enterprise GovCon procurement workstation. It unifies government contracting, business development, and commercial sales in one platform with automated solicitation discovery from SAM.gov, DLA, Unison, and Bonfire.

What does VPW stand for?

VPW stands for Viceroy Productivity Workstation — Viceroy NM's GovCon-focused enterprise platform powered by Trunnion AI agents.

What marketplaces does VPW pull solicitations from?

VPW imports solicitations daily from four federal and commercial marketplaces: SAM.gov, DLA (Defense Logistics Agency), Unison Marketplace, and Bonfire. Additional sources can be added as custom integrations.

Who is VPW for?

VPW is built for government contractors of all sizes — primes, subs, and small businesses pursuing federal, state, and commercial opportunities. It unifies business development, capture, proposal, and commercial sales workflows in a single GovCon ERP.

What is the VPW 15-state bid lifecycle?

VPW manages opportunities through 15 distinct lifecycle states from initial discovery through award and contract execution, with configurable approval gates at each stage. The lifecycle ensures no opportunity falls through the cracks and every bid follows your team's review process.

Does VPW match opportunities to vendors automatically?

Yes. VPW uses AI-powered vendor matching that scores opportunities by geography, NAICS codes, set-aside categories, past performance, SAM registration status, and capacity. Top matches surface to the BD team for review.

Does VPW handle line-item pricing?

Yes. VPW supports detailed line-item pricing with automated margin calculations, cost roll-ups, vendor quote tracking, and bid/no-bid analysis. Pricing logic is auditable and configurable per opportunity.

Can VPW handle commercial sales too?

Yes. VPW unifies GovCon procurement with traditional commercial sales pipelines so contractors don't need separate CRMs for federal and commercial business. Both pipelines share vendor relationships, customer data, and reporting.

Is VPW a CRM?

VPW includes CRM functionality (contacts, accounts, opportunities) but is broader than a traditional CRM. It's a unified GovCon ERP combining solicitation discovery, capture management, proposal workflow, vendor management, pricing, and commercial sales.

How does VPW use Trunnion AI?

VPW uses Trunnion AI agents for automated solicitation parsing, requirement extraction, vendor matching, pricing analysis, capability mapping, proposal drafting assistance, and compliance review. All agent actions are logged and reviewable.

Where can I learn more about VPW?

Visit trunnion.ai/vpw for the complete VPW overview, integrations list, and demo scheduling.

What is Cortrova?

Cortrova is Trunnion AI's manufacturing ERP product. It provides 205+ platform features across 17 operational departments with a 3-tier AI architecture: 51 department agents, 7 division aggregators, and 6 executive agents delivering real-time production intelligence.

Who is Cortrova for?

Cortrova is built for discrete and process manufacturers — small to mid-size manufacturing companies that need a unified ERP with embedded AI intelligence rather than legacy ERPs with AI bolt-ons.

What is Cortrova's 3-tier AI architecture?

Cortrova has 51 department-level AI agents handling specific functions (quality, scheduling, inventory, maintenance, etc.), 7 division aggregator agents that synthesize across related departments, and 6 executive agents that deliver C-level operational intelligence and strategic recommendations.

How many departments does Cortrova cover?

Cortrova covers 17 operational departments including production planning, quality, maintenance, inventory, procurement, sales, customer service, accounting, HR, engineering, R&D, shipping, receiving, IT, EHS, finance, and executive operations.

What is SQDI&F integration in Cortrova?

SQDI&F stands for Safety, Quality, Delivery, Inventory, and Finance — the five core operational metrics manufacturers track. Cortrova integrates SQDI&F across modules so data flows in unified streams rather than departmental silos. Each agent has access to relevant SQDI&F context.

Does Cortrova handle production scheduling?

Yes. Cortrova includes AI-powered production scheduling with capacity planning, constraint optimization, real-time rescheduling on disruption, and what-if scenario modeling.

Does Cortrova detect anomalies?

Yes. Cortrova provides real-time anomaly detection across production lines, quality processes, inventory movements, and equipment performance with automated escalation to the right operator or department.

Is Cortrova on-premises or cloud?

Cortrova supports both. Cloud deployments use AWS infrastructure with multi-tenant isolation. On-premises deployments support manufacturers with sensitive IP, sovereignty requirements, or limited connectivity. Hybrid deployments are also available.

How does Cortrova compare to SAP?

SAP is a legacy ERP with decades of feature accumulation, very high cost, multi-year implementation, and AI added recently as a separate module. Cortrova is AI-native — agents are foundational, not bolt-on. It's purpose-built for discrete manufacturing, deploys in months not years, and costs a fraction of SAP.

What's Cortrova's website?

Cortrova has its own product site at cortrova.com, with the company overview at trunnion.ai/cortrova.

What is Haultro?

Haultro is Trunnion AI's waste management and fleet optimization SaaS for waste haulers, junk removal companies, dumpster rental operators, and roll-off services. It includes IoT bin monitoring, AI route optimization, real-time fleet tracking, SLA compliance monitoring, and a 39-agent Trunnion AI architecture.

Who is Haultro for?

Haultro is built for waste management operators of all sizes — from small junk removal companies managing 25 bins to enterprise franchise operators managing 50,000+ bins across multiple regions.

Does Haultro support IoT bin monitoring?

Yes. Haultro integrates with IoT bin sensors for real-time fill-level telemetry. Operators can see which bins are full, empty, or contaminated without sending a truck to check. This drives smarter routing and prevents missed pickups.

How does Haultro optimize routes?

Haultro uses AI-powered route optimization that considers bin fill levels, traffic, vehicle capacity, driver schedules, customer SLAs, and historical patterns. Routes update dynamically as new data comes in. Drivers see optimized routes on their mobile devices.

Does Haultro track SLA compliance?

Yes. Haultro monitors service level agreements per customer with automated escalation when an SLA is at risk. Customers get pickup confirmations; managers get exception reports; missed pickups trigger immediate workflow remediation.

What is Trunnion Vision AI in Haultro?

Trunnion Vision AI is the computer vision capability built into Haultro. It analyzes bin contents from photos taken by drivers or fixed cameras to detect contamination, classify waste types, identify hazardous materials, and validate service completion.

How many AI agents does Haultro use?

Haultro uses a 39-agent Trunnion AI architecture spanning operations (dispatch, routing, monitoring), fleet (vehicle health, driver performance, maintenance), and finance (invoicing, collections, route profitability).

Is Haultro multi-tenant?

Yes. Haultro is multi-tenant SaaS supporting independent operators on shared infrastructure with strict data isolation. Enterprise franchise operators can also deploy single-tenant instances if required.

What's Haultro's website?

Haultro has its own product site at haultro.com, with the company overview at trunnion.ai/haultro.

What is ZennVue?

ZennVue is Trunnion AI's full-stack ERP and two-sided marketplace for the creative events industry. It serves photographers, videographers, DJs, event planners, caterers, and venues with 12 functional domains and 25+ AI features.

Who is ZennVue for?

ZennVue is for creative event vendors — wedding photographers, videographers, DJs, event planners, caterers, florists, venues — and the couples or event hosts who hire them. It serves both sides of the marketplace with role-specific tools.

What functional domains does ZennVue cover?

ZennVue covers 12 domains: CRM, lead management, invoicing, contracts, e-signatures, gallery delivery, scheduling, calendar sync, email/SMS automation, payment processing, vendor portal, and client portal.

Does ZennVue have AI features?

Yes. ZennVue includes 25+ AI features powered by Trunnion AI multi-LLM orchestration, including AI proposal drafting, contract review, smart inquiry responses, gallery curation, post-event automation, and personalized client communication.

Is ZennVue a marketplace?

Yes. ZennVue is a two-sided marketplace where couples and event hosts find and book vendors, while vendors run their entire business through the same platform — eliminating the gap between marketplace platforms (HoneyBook, The Knot) and operations software.

Does ZennVue replace HoneyBook or Dubsado?

For most creative event vendors, yes. ZennVue includes the CRM, contracts, invoicing, and client communication features of HoneyBook or Dubsado, plus a two-sided marketplace, AI features, and integrated portals — at competitive pricing.

What's the market opportunity for ZennVue?

The creative events and weddings market is projected to reach $52B by 2034. ZennVue is purpose-built to serve the vendor and client segments of this growing market with AI-native operations software.

Does ZennVue support real-time communication?

Yes. ZennVue uses WebSocket-based real-time communication across all platform events — message threads, booking updates, payment notifications, gallery access, and inquiry responses all happen in real-time without page refreshes.

What's ZennVue's website?

ZennVue has its own product site at zennvue.com, with the company overview at trunnion.ai/zennvue.

What is RealtrAI?

RealtrAI is Trunnion AI's suite of 17 AI-powered tools built specifically for real estate agents. Tools include luxury listing description generators, vacation rental ROI calculators, investor pitch generators, and smart 5-email follow-up sequences.

How many AI tools does RealtrAI have?

RealtrAI includes 17 AI-powered tools covering listing creation, marketing copy, investor pitches, vacation rental analysis, neighborhood content, photography planning, lead nurturing, and follow-up automation.

Who is RealtrAI for?

RealtrAI is built for individual real estate agents and brokerages who want enterprise-grade AI in their workflow. It's especially powerful for luxury agents, investor-focused agents, and agents managing vacation rental portfolios.

What is the Luxury Story Builder in RealtrAI?

Luxury Story Builder is a RealtrAI tool that generates premium listing descriptions with neighborhood-specific language, evocative property storytelling, and matching photography shot lists. It's built specifically for luxury price points where generic listing copy fails.

What is the Vacation Rental ROI tool in RealtrAI?

The Vacation Rental ROI tool projects investment returns for short-term rental properties using occupancy benchmarks, ADR comparables, expense modeling, and Airbnb vs VRBO platform comparison. Agents use it to advise investor clients.

Can RealtrAI write investor pitches?

Yes. The Investor Pitch tool generates customized investment opportunity pitches for investor clients based on the property profile, market data, expected returns, and the investor's stated criteria.

Does RealtrAI handle lead follow-up?

Yes. RealtrAI generates smart 5-email drip sequences personalized per lead, with automated send scheduling and response detection. Sequences adapt based on lead engagement signals.

What is the Property Workspace in RealtrAI?

Property Workspace is a per-listing dashboard in RealtrAI that tracks status, links to all marketing platforms (MLS, Zillow, Redfin, Compass, etc.), runs AI tools for that specific property, and centralizes the agent's work on each listing.

What's RealtrAI's website?

RealtrAI has its own product site at realtrai.com, with the company overview at trunnion.ai/realtrai.

What is FizzSense?

FizzSense is Trunnion AI's vertically integrated IoT SaaS platform for CO2 tank logistics serving beverage distributors and restaurant networks. It uses cellular IoT telemetry for predictive depletion forecasting, automated demand queues, and AI-powered fleet route optimization.

Who is FizzSense for?

FizzSense serves two sides of the CO2 supply chain: restaurants and bars that need reliable CO2 for beverage carbonation, and the beverage distributors that supply them with CO2 tanks.

What problem does FizzSense solve?

CO2 tank distribution today is reactive — distributors run blind routes and restaurants frequently run out, losing revenue. FizzSense uses IoT sensors to forecast depletion and trigger refills before runouts, eliminating blind routes and preventing costly emergencies.

How does FizzSense use IoT?

FizzSense smart monitoring scales attach to CO2 tanks and report fill levels via cellular telemetry. Each scale streams real-time data to the FizzSense platform where predictive models forecast when each tank will reach a refill threshold.

Does FizzSense optimize delivery routes?

Yes. FizzSense uses AI-powered fleet route optimization for distributors. Routes are built dynamically from the predictive demand queue rather than fixed schedules, eliminating empty miles and prioritizing imminent runout risks.

What are FizzSense's economics?

FizzSense targets 80%+ SaaS margins with deterministic analytics. Revenue comes from per-restaurant subscription fees, distributor platform fees, and IoT scale leasing. Unit economics are strong because the IoT scale captures recurring data value over multi-year deployment.

When will FizzSense be available?

FizzSense is in pre-launch with early access available. Visit trunnion.ai/fizzsense to request early access for your distribution business or restaurant network.

How many products does Trunnion AI power?

Trunnion AI currently powers seven products: DAF, VPW, Cortrova, Haultro, ZennVue, RealtrAI, and FizzSense. Each is a vertical SaaS product purpose-built for a specific industry.

Are all Trunnion AI products live?

Six of the seven are live in production: DAF, VPW, Cortrova, Haultro, ZennVue, and RealtrAI. FizzSense is in pre-launch with early access.

Can products from different verticals integrate?

Yes. Because all Trunnion-powered products run on the same orchestration core, cross-product integrations are straightforward. For example, a Cortrova manufacturing customer also using VPW for GovCon procurement gets unified data flow between the two systems.

Will Trunnion AI add more products?

Yes. Viceroy NM continues to launch new vertical SaaS products on the Trunnion AI orchestration framework. New verticals are typically chosen based on market gaps where governance, compliance, and AI agent orchestration deliver clear ROI.

Can I license just the Trunnion AI core framework?

Yes. Enterprise customers can license the underlying Trunnion AI orchestration framework directly to build custom internal applications. Contact Viceroy NM for framework licensing and architecture support options.

How does Trunnion AI serve the defense industry?

Trunnion AI serves defense through DAF (Declarative Agentic Framework) — a multi-agent orchestration platform purpose-built for DoD and intelligence community needs: classification enforcement, CAC/PIV authentication, air-gap deployment, NIST 800-53 compliance, and 150+ pre-built mission agents.

What is the best AI platform for DoD contractors?

For DoD contractors needing governed multi-agent AI with classified-environment support, DAF (Trunnion AI's defense product) is purpose-built for this use case. It's air-gap deployable, NIST 800-53 mapped, CAC/PIV-enabled, and runs at 5–10x lower cost than incumbent defense AI platforms.

Can AI run in a SCIF?

Yes — with the right architecture. AI can run in a Sensitive Compartmented Information Facility (SCIF) using locally-hosted open-source LLMs (e.g., via Ollama) with no external network connectivity. Trunnion AI's DAF product is architected specifically for SCIF and air-gapped classified environments.

What AI platforms support TS/SCI?

Few commercial AI platforms support TS/SCI environments because they require air-gap deployment, classification enforcement, and CAC/PIV identity. Trunnion AI's DAF product is designed for TS/SCI, with five-level classification enforcement architecturally built into the platform.

How does AI help defense intelligence analysis?

AI agents can ingest large volumes of unstructured intelligence (cables, reports, signals data, imagery analysis), summarize key findings, identify patterns and anomalies, draft analytical assessments, and support analysts in tradecraft. Governed agentic AI like Trunnion AI maintains audit trails and HITL controls throughout.

What is AI mission planning?

AI mission planning uses agents to analyze terrain, weather, threat data, asset availability, and mission requirements to generate planning options. Human commanders evaluate options, request modifications, and approve final plans. The AI accelerates analysis; humans retain command authority.

Can Trunnion AI replace JADC2?

No — JADC2 (Joint All-Domain Command and Control) is a DoD strategic concept, not a single product. Trunnion AI can serve as an orchestration layer within a JADC2-aligned architecture, coordinating agents across data sources and decision systems with the governance JADC2 requires.

What is Mosaic Warfare and does Trunnion AI fit?

Mosaic Warfare is a DARPA concept for distributed, composable kill chains. Trunnion AI's multi-agent architecture aligns naturally with Mosaic principles — small, specialized AI agents that compose into mission workflows, with governance and audit at every layer.

How does AI help government contractors?

AI helps GovCon companies automate solicitation discovery, requirement extraction, capability mapping, vendor matching, pricing analysis, proposal drafting, and compliance review — capturing more opportunities with less manual labor. Trunnion AI's VPW product is built specifically for these workflows.

Can AI write a federal proposal?

AI can draft sections of a federal proposal — past performance summaries, capability statements, technical approach narratives — based on the contractor's existing knowledge base and the solicitation requirements. Final review and submission remain with humans. Trunnion AI's VPW supports this workflow with HITL gates.

What is the best CRM for GovCon?

Generic CRMs (Salesforce, HubSpot) lack GovCon-specific features. Purpose-built GovCon platforms like VPW (Trunnion AI) handle SAM.gov solicitation imports, NAICS-based matching, set-aside qualification, line-item pricing, 15-state bid lifecycle, and unified federal+commercial pipelines.

How do I find federal contracting opportunities?

Federal opportunities are posted on SAM.gov and other marketplaces (DLA, Unison, Bonfire). Manually monitoring these is time-consuming. Tools like Trunnion AI's VPW automatically import opportunities daily and use AI to match them to your capability profile.

What is shipping tracking for federal contracts?

For federal contracts requiring delivery, tracking includes proof of delivery, compliance with packaging and shipping requirements, FAR-mandated reporting, and timeline accountability. VPW handles these as part of the contract execution workflow.

What is AI-powered manufacturing ERP?

AI-powered manufacturing ERP embeds AI agents into core ERP functions — production scheduling, quality, inventory, maintenance — rather than offering AI as a separate analytics module. Cortrova (Trunnion AI) is an example: 51 department-level AI agents integrated across all manufacturing operations.

How does AI help small manufacturers?

AI helps small manufacturers automate scheduling decisions, detect quality anomalies before they become defects, optimize inventory, predict maintenance needs, and surface real-time operational intelligence — capabilities historically reserved for enterprises with massive data science teams.

What is real-time production intelligence?

Real-time production intelligence is continuous AI-driven monitoring of all factory operations with anomaly detection, predictive insights, and automated escalation. Operators see issues as they emerge, not after end-of-day reports. Cortrova provides this across 17 departments.

Can AI replace SAP for manufacturers?

For small and mid-size manufacturers, AI-native ERPs like Cortrova can replace legacy SAP at a fraction of the cost and implementation time. Large enterprises with complex SAP customizations typically migrate gradually rather than wholesale.

What is Industry 4.0 and how does AI fit?

Industry 4.0 refers to the integration of IoT, AI, robotics, and connected systems in manufacturing. AI agents are central to Industry 4.0 — they coordinate decisions across IoT sensors, MES, ERP, and human operators. Cortrova is purpose-built for Industry 4.0 manufacturing operations.

How does AI help real estate agents?

AI helps real estate agents write better listing descriptions, generate investor pitches, analyze vacation rental ROI, draft personalized follow-up sequences, and create luxury listing narratives. RealtrAI provides 17 such tools purpose-built for agents.

Will AI replace real estate agents?

No. AI handles repetitive copywriting, analysis, and follow-up tasks. Real estate transactions still require human relationships, negotiation, and trust. AI tools like RealtrAI free agents to spend more time on relationship-driven work.

How can AI help with vacation rentals?

AI can analyze vacation rental ROI using occupancy benchmarks, ADR comparables, and expense modeling; write platform-optimized listings for Airbnb and VRBO; and generate dynamic pricing recommendations. RealtrAI's Vacation Rental ROI tool handles these analyses.

What is luxury real estate marketing?

Luxury real estate marketing emphasizes lifestyle storytelling, neighborhood-specific narratives, and visually-driven campaigns that distinguish premium properties from generic listings. AI tools like RealtrAI's Luxury Story Builder generate copy and shot lists at this elevated standard.

What is the best ERP for wedding photographers?

Wedding photographers need a unified system for inquiries, contracts, invoicing, gallery delivery, and client communication. ZennVue is purpose-built for this — combining the operations features of HoneyBook or Dubsado with a vendor marketplace and integrated AI.

How is the wedding industry changing with AI?

AI is reshaping how creative event vendors handle inquiry response, proposal drafting, contract review, gallery curation, and client communication — automating busywork so vendors focus on creative work and relationships. Two-sided marketplaces like ZennVue compress the booking cycle from weeks to days.

What is a creative events ERP?

A creative events ERP is a unified operations platform for wedding and event vendors covering CRM, contracts, invoicing, scheduling, gallery delivery, and client communication. Unlike generic CRMs, creative events ERPs understand vendor-specific workflows. ZennVue is the leading AI-native creative events ERP.

How does AI help waste management companies?

AI helps waste management with route optimization, IoT bin monitoring, SLA compliance, contamination detection (computer vision), driver dispatch, and financial management. Haultro (Trunnion AI) is the leading AI-native waste management platform.

What is smart waste management?

Smart waste management uses IoT sensors on bins to report fill levels in real time, AI route optimization to dispatch trucks only when needed, and computer vision to detect contamination. The result is fewer trucks, fewer missed pickups, and lower operating costs.

How do IoT bin sensors work?

IoT bin sensors are devices attached to waste bins that measure fill level (typically with ultrasonic or weight sensing) and report telemetry over cellular networks. Platforms like Haultro ingest the telemetry and use AI to forecast pickups and optimize routes.

How does AI improve fleet logistics?

AI improves fleet logistics through predictive demand forecasting (when each customer needs delivery), dynamic route optimization (continuously updated as new data arrives), driver dispatch automation, and fuel/maintenance prediction. Trunnion AI's Haultro and FizzSense both apply these principles.

What is predictive depletion forecasting?

Predictive depletion forecasting uses IoT sensor data and AI models to predict when a consumable (CO2 in restaurants, waste in bins) will reach a refill threshold. This enables proactive delivery rather than reactive emergency runs. FizzSense applies it to CO2 logistics.

What is the CO2 distribution business?

CO2 distribution is the supply chain that delivers compressed CO2 tanks to restaurants, bars, and food service operations for beverage carbonation, draft beer dispensing, and food preservation. Distributors operate fleets serving regional customer networks. FizzSense modernizes this industry with IoT and AI.

How much does Trunnion AI cost?

Trunnion AI uses custom enterprise pricing scoped to each deployment. Pricing depends on agent count, classification ceiling, deployment model (cloud, edge, or air-gap), and support requirements. Contact Viceroy NM at trunnion.ai/contact for a quote tailored to your environment.

Why doesn't Trunnion AI have public pricing?

Enterprise and defense AI deployments vary widely — agent count, classification ceiling, deployment environment, integration complexity, and support level all affect cost. Public pricing would either oversimplify (and fail customers) or overstate (and scare them off). Custom pricing produces better outcomes.

Are there free or trial versions of Trunnion AI?

Trunnion AI itself doesn't offer a free tier — it's enterprise infrastructure. However, Trunnion-powered products like RealtrAI offer trials and entry tiers for individual users. Contact Viceroy NM for product-specific trial options.

How does Trunnion AI compare on cost vs Palantir?

Trunnion AI's defense product DAF runs at approximately 5–10x lower total cost than Palantir Foundry deployments for comparable use cases. Cost savings come from a focused multi-agent architecture, faster deployment, lower professional services overhead, and LLM-agnostic licensing.

What's included in a Trunnion AI license?

Standard licenses include the orchestration framework, configured agent libraries, deployment support, environment-specific compliance mapping, dedicated success engineering, and SLA-backed support. Custom agent development and on-site deployment support are available as add-ons.

Are there discounts for multi-product customers?

Yes. Customers using multiple Trunnion-powered products (e.g., Cortrova for manufacturing + VPW for GovCon) typically receive bundled pricing reflecting the shared platform foundation. Contact sales for multi-product quotes.

What deployment models does Trunnion AI support?

Cloud (AWS-native multi-tenant or single-tenant), on-premises (single-tenant in customer infrastructure), edge (distributed deployment near data sources), and air-gap (fully isolated, no external connectivity). Hybrid combinations are also supported.

Can Trunnion AI run in AWS GovCloud?

Yes. Trunnion AI is AWS-native and deploys in AWS GovCloud (US) for federal customers requiring FedRAMP-aligned infrastructure. Other AWS regions including standard commercial regions are also supported.

Does Trunnion AI run on Azure or Google Cloud?

Trunnion AI's primary deployment target is AWS, but the architecture is cloud-portable. Azure and Google Cloud deployments are supported as enterprise engagements with dedicated environment setup. Contact Viceroy NM for non-AWS deployment timelines.

Can I self-host Trunnion AI?

Yes. Self-hosted on-premises deployments are fully supported and common for defense, manufacturing, and customers with sovereignty requirements. Self-hosted deployments include local LLM hosting via Ollama for full data isolation.

What's the difference between cloud and air-gap deployment?

Cloud deployments run in AWS (commercial or GovCloud) with external network connectivity for SaaS LLM APIs. Air-gap deployments run in isolated customer environments with no external connectivity, using locally-hosted open-source LLMs via Ollama. Air-gap is required for classified or sovereignty-sensitive workloads.

Can I switch deployment models later?

Yes. Trunnion AI supports migration between deployment models — for example, starting with multi-tenant cloud and migrating to single-tenant or on-premises as the deployment matures. Migration typically requires planning but doesn't require re-architecting workflows.

How long does Trunnion AI take to deploy?

A typical Trunnion AI deployment follows four phases: discovery and scoping (2–4 weeks), supervised pilot (4–8 weeks), progressive autonomy (8–12 weeks), and scale optimization. Most enterprise deployments reach production within 90 days; defense air-gap deployments may take longer based on facility access.

What is the deployment process?

Phase 1: discovery — map systems, identify automation candidates, define governance. Phase 2: supervised pilot — deploy with HITL gates on every action, full observability. Phase 3: progressive autonomy — transition proven workflows to semi-autonomous. Phase 4: scale and optimize — expand across departments with continuous monitoring.

Do I need data scientists to deploy Trunnion AI?

No. Trunnion AI is enterprise infrastructure designed to be deployed and operated by IT teams and business analysts, not ML engineers. Pre-built agents handle most workflows; custom agent development is supported but not required.

What does the supervised pilot phase involve?

In the supervised pilot, agents run with HITL approval gates on every meaningful action. Operators see exactly what each agent wants to do, approve or reject, and provide feedback. This builds trust, surfaces edge cases, and produces audit-ready validation before transitioning to higher autonomy.

When can I trust agents to run autonomously?

Autonomy is granted progressively, workflow by workflow, based on observed performance during the pilot. Routine, lower-consequence workflows transition first. High-consequence workflows retain HITL gates indefinitely. The decision to grant autonomy is always the customer's, not Trunnion's.

What kind of support does Trunnion AI provide?

Standard licenses include dedicated success engineering during deployment, ongoing technical support with SLA-backed response times, regular platform updates, and access to the customer success team. On-site deployment support and 24/7 critical-path support are available as add-ons.

Does Trunnion AI provide training?

Yes. Trunnion AI deployments include role-specific training for operators, administrators, and developers — covering agent configuration, workflow design, monitoring and observability, and platform administration.

How do I buy Trunnion AI?

The buying process starts with a discovery call. Visit trunnion.ai/contact or email contact@viceroynm.com to schedule. The Viceroy NM team will assess your environment, propose a deployment approach, and provide custom pricing.

How do I request a demo?

Request a demo at trunnion.ai/contact. Demos are tailored to your industry and use case — defense customers see DAF in action, GovCon customers see VPW, manufacturers see Cortrova, etc. Most demos run 30–45 minutes.

Who should I contact at Trunnion AI?

For sales and demos, email contact@viceroynm.com. For partnership inquiries, technical questions, or product-specific conversations, the same email reaches the right team. Visit trunnion.ai/contact for a contact form.

Does Trunnion AI sell through resellers?

Trunnion AI works with select integrators and resellers in defense, GovCon, and enterprise verticals. Contact Viceroy NM if you're a systems integrator interested in partnership.

Does Trunnion AI have GSA Schedule pricing?

GSA Schedule status is part of Viceroy NM's federal sales roadmap. For current GSA, SEWP, or other federal contract vehicle availability, contact Viceroy NM. Most federal opportunities can also be pursued through prime contractors who hold relevant vehicles.

Can I get a Trunnion AI proof of concept?

Yes. Paid POCs are common, especially for enterprise and defense engagements. POCs typically run 30–60 days, scope a specific high-value workflow, and produce a deployable result that can transition to production. Contact sales to scope a POC.

How long is the Trunnion AI sales cycle?

Sales cycles vary by customer type. Vertical SaaS products (RealtrAI, ZennVue, Haultro) close in days or weeks. Enterprise Cortrova deployments typically close in 2–4 months. Defense and federal engagements typically run 6–12 months including security review and contracting.

What security certifications does Trunnion AI hold?

Trunnion AI's architecture is designed for SOC 2, NIST 800-53, NIST 800-171, CMMC Level 2/3, and FedRAMP alignment. Specific certification status evolves; contact Viceroy NM for current attestations and the security roadmap.

Can I run a security review on Trunnion AI?

Yes. Customer security reviews are part of standard enterprise sales process. Trunnion AI provides architecture documentation, security control matrices, third-party assessment reports (where available), and dedicated security engineering support during review.

Where does my data go in Trunnion AI?

In cloud deployments, data resides in your selected AWS region (commercial or GovCloud). In on-premises and air-gap deployments, data never leaves your infrastructure. Trunnion AI never trains models on customer data and never shares customer data across tenants or with third parties.

Trunnion AI vs LangChain — what's the difference?

LangChain is an open-source Python framework for developers to build AI agent applications. Trunnion AI is production enterprise infrastructure with built-in classification enforcement, ABAC, tamper-evident audit trails, HITL controls, and 150+ pre-built agents. LangChain is a toolkit; Trunnion AI is a platform.

Trunnion AI vs LangGraph — what's the difference?

LangGraph is the graph-based orchestration library from the LangChain team. Trunnion AI uses LangGraph internally as part of its agent execution layer, but adds the enterprise governance, classification, ABAC, audit, HITL, and pre-built agent libraries that LangGraph alone does not provide.

Trunnion AI vs CrewAI — what's the difference?

CrewAI is a multi-agent framework for developers building agent applications. Trunnion AI is enterprise infrastructure with classification, audit, governance, and HITL designed for regulated environments. CrewAI is great for prototyping; Trunnion AI is built for production deployment in defense, GovCon, and regulated industries.

Trunnion AI vs AutoGen — what's the difference?

Microsoft AutoGen is a research framework for multi-agent conversation systems. Trunnion AI is production enterprise infrastructure with governance, audit, and compliance. AutoGen suits research and prototyping; Trunnion AI suits regulated production deployment.

Trunnion AI vs OpenAI Assistants API — what's the difference?

OpenAI Assistants API provides single-agent capabilities tied to OpenAI models. Trunnion AI is LLM-agnostic, multi-agent, and includes enterprise governance OpenAI doesn't provide — classification enforcement, ABAC, tamper-evident audit, HITL controls, air-gap deployment.

Trunnion AI vs Anthropic Computer Use — what's the difference?

Anthropic's Computer Use enables Claude to control a computer interface. Trunnion AI orchestrates multi-agent workflows that may include computer-use capabilities. They're complementary — Trunnion AI can deploy Claude with Computer Use as one tool in a broader governed workflow with audit trails and HITL gates.

Trunnion AI vs Palantir Foundry — what's the difference?

Palantir Foundry is a data integration and ontology platform with AI features layered on. Trunnion AI's DAF product is purpose-built multi-agent orchestration. DAF runs at 5–10x lower cost, deploys in weeks rather than months, supports any LLM, and provides 150+ pre-built defense mission agents.

Trunnion AI vs Anduril Lattice — what's the difference?

Anduril Lattice is an autonomous defense operating system focused on edge sensor fusion and unmanned systems control. Trunnion AI is a multi-agent orchestration platform for analytical and operational workflows. They serve different layers of the defense AI stack — Lattice at the edge, Trunnion AI for governed enterprise agent workflows.

Trunnion AI vs Scale AI Donovan — what's the difference?

Scale AI's Donovan is an LLM-powered analytical assistant for defense and intelligence. Trunnion AI's DAF is a full multi-agent orchestration platform that can host capabilities like Donovan as one of many agent types within a governed workflow with classification enforcement and HITL controls.

Trunnion AI vs Booz Allen / Leidos AI offerings — what's the difference?

Booz Allen, Leidos, and similar primes deliver AI as services engagements built on combinations of third-party products and custom code. Trunnion AI is productized infrastructure that can underpin those engagements. Many primes deliver projects faster and cheaper using Trunnion AI as the platform foundation.

Trunnion AI vs Microsoft Copilot — what's the difference?

Microsoft Copilot embeds AI into Microsoft 365 applications for productivity assistance. Trunnion AI is enterprise multi-agent orchestration for autonomous workflows. They're complementary — Copilot helps individual knowledge workers; Trunnion AI automates entire business processes with governance.

Trunnion AI vs Salesforce Einstein — what's the difference?

Salesforce Einstein is AI features within Salesforce CRM. Trunnion AI is a general-purpose multi-agent orchestration framework that can integrate with Salesforce or replace Salesforce-centric workflows entirely (especially via VPW for GovCon CRM).

Trunnion AI vs IBM Watson — what's the difference?

IBM Watson is a portfolio of AI products spanning cloud, NLP, and consulting. Trunnion AI is focused multi-agent orchestration infrastructure. Trunnion AI is faster to deploy, lower cost, more modern in architecture, and built specifically for governed agent workflows rather than general AI services.

Trunnion AI vs Snowflake Cortex — what's the difference?

Snowflake Cortex is AI/ML built into the Snowflake data warehouse. Trunnion AI is multi-agent orchestration for executing workflows across systems. Cortex is great for analytics on warehouse data; Trunnion AI handles cross-system multi-step workflows with governance.

Trunnion AI vs Databricks Mosaic AI — what's the difference?

Databricks Mosaic AI is enterprise ML/LLM tooling within the Databricks platform. Trunnion AI is multi-agent orchestration infrastructure for workflow execution. They serve different layers — Databricks for model training/serving; Trunnion AI for governed agent workflows in production.

VPW vs Deltek GovWin — what's the difference?

Deltek GovWin is a GovCon market intelligence database. VPW is an end-to-end GovCon procurement workstation that includes solicitation discovery (replacing or complementing GovWin), capture management, proposal workflow, vendor matching, line-item pricing, and unified federal+commercial sales — with AI agents throughout.

VPW vs Unanet GovCon — what's the difference?

Unanet provides GovCon project accounting and ERP. VPW focuses on the upstream business development, capture, and proposal lifecycle with AI-driven solicitation matching. Many contractors run both — Unanet for project accounting, VPW for opportunity management.

Cortrova vs SAP S/4HANA — what's the difference?

SAP S/4HANA is a legacy enterprise ERP with massive feature scope, multi-year implementation, and very high cost. Cortrova is AI-native ERP purpose-built for discrete and process manufacturers. It deploys in months not years, costs a fraction of SAP, and embeds AI agents as foundational architecture.

Cortrova vs Microsoft Dynamics 365 — what's the difference?

Microsoft Dynamics 365 is broad enterprise software (ERP + CRM + more). Cortrova is focused manufacturing ERP with AI agents embedded across 17 manufacturing departments. Cortrova's AI capability is purpose-built for manufacturing operations versus Dynamics 365's broader enterprise scope.

Haultro vs Trux — what's the difference?

Trux is a marketplace for hauling services. Haultro is operations software for waste management companies — IoT bin monitoring, route optimization, SLA compliance, fleet tracking — purpose-built around AI agents. They serve different needs in waste/hauling.

Haultro vs AMCS Platform — what's the difference?

AMCS is a long-established waste management software vendor with enterprise scale and legacy architecture. Haultro is AI-native, modern cloud SaaS purpose-built for the IoT and AI era — faster deployment, lower cost, embedded AI agents across operations, fleet, and finance.

ZennVue vs HoneyBook — what's the difference?

HoneyBook is a CRM and operations platform for creative event vendors. ZennVue includes equivalent operations features plus a two-sided marketplace (where couples discover vendors), 25+ AI features, and real-time WebSocket communication. ZennVue is broader and more AI-native.

ZennVue vs The Knot Pro — what's the difference?

The Knot Pro is primarily a marketplace for wedding vendors. ZennVue combines marketplace with full vendor operations (CRM, contracts, invoicing, gallery delivery) so vendors don't need separate marketplace and operations tools. ZennVue's AI features streamline both sides.

RealtrAI vs CINC — what's the difference?

CINC is a real estate CRM and lead generation platform. RealtrAI is a suite of 17 AI-powered agent productivity tools — listing copy, investor pitches, vacation rental ROI, follow-up sequences. Many agents use both: CINC for CRM, RealtrAI for AI-driven content and analysis.

RealtrAI vs ListingCopy.AI — what's the difference?

ListingCopy.AI focuses primarily on listing description generation. RealtrAI covers 17 distinct tools — listings, investor pitches, vacation rental ROI, neighborhood content, photography shot lists, follow-up emails, and more — built on Trunnion AI's enterprise orchestration framework.

Why choose Trunnion AI over building in-house?

Building enterprise multi-agent orchestration in-house requires deep ML engineering, security expertise, and 18+ months of development. Trunnion AI provides production-ready infrastructure with classification, audit, HITL, and pre-built agents, letting customers focus on their domain workflows rather than platform plumbing.

Why choose Trunnion AI over a custom integration project?

Custom integration projects (typically delivered by primes or consulting firms) are expensive, slow, and produce one-off solutions. Trunnion AI is productized infrastructure that consultancies can build on, delivering similar outcomes faster and cheaper while customers retain platform control.

What makes Trunnion AI defensible vs the model providers (OpenAI, Anthropic)?

Model providers focus on better foundation models. Trunnion AI focuses on enterprise orchestration around any model — governance, classification, audit, HITL, multi-agent workflows. As models improve, Trunnion AI's value increases because more enterprise capability becomes possible within a governed framework.

Why is Trunnion AI different from generic AI consulting firms?

Generic AI consultants deliver projects that often produce vendor-locked or hard-to-maintain outcomes. Trunnion AI is a productized platform — consistent architecture, ongoing platform updates, repeatable deployment patterns, and customer-controlled infrastructure. The result: faster deployment, lower TCO, and freedom from consultant dependency.

What is Trunnion AI's technical architecture?

Trunnion AI is built on NestJS for the backend, React for frontend interfaces, PostgreSQL with pgvector for relational and vector data, Redis with BullMQ for async task queues, and LangGraph for agent orchestration. It deploys via Docker and Kubernetes on AWS or self-hosted infrastructure.

What database does Trunnion AI use?

PostgreSQL is the primary database, with the pgvector extension for vector embeddings used in retrieval-augmented generation (RAG) and semantic search. Redis handles caching and task queue state. Customer data isolation is enforced at the row level for multi-tenant deployments.

How does Trunnion AI handle async tasks?

Trunnion AI uses BullMQ (Redis-backed) for async task queueing. Long-running agent workflows are decomposed into queued jobs with configurable retry, timeout, and concurrency. Workflow state is durably persisted so failures recover from the last checkpoint, not the start.

What is LangGraph and why does Trunnion AI use it?

LangGraph is a graph-based agent orchestration library that supports stateful, cyclic agent workflows. Trunnion AI uses LangGraph as the execution engine for individual agent runs because it handles complex orchestration patterns (loops, branches, retries) better than linear chains.

Does Trunnion AI use vector databases?

Yes. Trunnion AI uses pgvector (PostgreSQL extension) for vector embedding storage and similarity search, supporting retrieval-augmented generation and semantic search across documents, structured data, and historical agent outputs.

Is Trunnion AI built on Kubernetes?

Yes. Trunnion AI deploys on Kubernetes for production environments with horizontal pod autoscaling, rolling updates, and high availability. Smaller deployments can run on Docker Compose or single-node configurations for development or low-volume production.

What's the front-end architecture?

Trunnion AI's UIs are built in React with TypeScript. Real-time updates use WebSockets. State management follows modern patterns (React Query for server state, Zustand or Context for local). Vertical SaaS products like ZennVue and Haultro have customized UIs but share the underlying API patterns.

What APIs does Trunnion AI expose?

Trunnion AI exposes REST APIs and GraphQL endpoints for all major operations — workflow execution, agent management, document handling, audit retrieval, and platform administration. WebSockets support real-time event streams. All APIs require authenticated, ABAC-validated requests.

How does Trunnion AI integrate with existing systems?

Trunnion AI integrates via REST/GraphQL APIs (outbound and inbound), webhooks, message queues, and direct database connectors. Pre-built integrations exist for SAM.gov, Google Workspace, Microsoft 365, Stripe, Salesforce, and major enterprise systems. Custom integrations are part of standard deployment.

Does Trunnion AI support webhooks?

Yes. Inbound webhooks trigger agent workflows from external systems. Outbound webhooks notify external systems of agent results, workflow state changes, or audit events. All webhook traffic is authenticated and signed.

Can I use my own LLM with Trunnion AI?

Yes. Trunnion AI is LLM-agnostic. Bring your own OpenAI key, Anthropic key, Google key, Mistral key, or Ollama-hosted local model. Many enterprise customers use multiple providers in the same deployment, routing different agent roles to different models based on cost, capability, or compliance requirements.

Does Trunnion AI support custom models?

Yes. Customers with proprietary or fine-tuned models can deploy them locally via Ollama or via OpenAI-compatible APIs. Trunnion AI treats them like any other LLM provider — agents can be routed to custom models per workflow requirement.

Does Trunnion AI support MCP (Model Context Protocol)?

Trunnion AI's tool authorization framework is compatible with MCP-style tool servers. Customers can wrap existing MCP tools as Trunnion AI executor agents to expose them to multi-agent workflows with full ABAC and audit. Native MCP integration is on the roadmap.

How fast is Trunnion AI?

Performance depends on the underlying LLMs and tool calls in each workflow. Agent orchestration overhead is typically under 100ms per step. Most workflow latency is dominated by LLM inference time and external API calls, not orchestration.

How does Trunnion AI scale?

Trunnion AI scales horizontally — workers process queued tasks in parallel across multiple Kubernetes pods. PostgreSQL and Redis scale with standard cloud database approaches (read replicas, managed services). Multi-tenant deployments share infrastructure with per-tenant isolation; single-tenant deployments scale independently.

What's Trunnion AI's uptime SLA?

Standard SaaS SLAs target 99.9% uptime with multi-region failover for hosted deployments. Enterprise contracts can specify higher SLAs (99.95% or 99.99%) with corresponding architectural and operational requirements. Air-gap deployments have customer-defined uptime targets.

How many agents can run concurrently?

Concurrent agent capacity depends on infrastructure provisioning. A standard production deployment supports hundreds of concurrent agent workflows. Larger deployments scale horizontally — additional worker pods increase concurrency linearly.

How does Trunnion AI handle authentication?

Authentication options include OAuth/OIDC, SAML 2.0 SSO, CAC/PIV smart cards (for federal), and API key authentication for service-to-service calls. Multi-factor authentication is supported throughout. All authentication events feed the audit trail.

Does Trunnion AI support SSO?

Yes. Standard OIDC and SAML 2.0 SSO are supported, integrating with Okta, Azure AD, Google Workspace, Auth0, Ping, and any standards-compliant identity provider. Federated identity attributes flow into the ABAC policy engine for consistent enforcement.

How is data encrypted in Trunnion AI?

Data is encrypted in transit (TLS 1.3) and at rest (AES-256). Database encryption uses cloud-native key management (AWS KMS or customer-managed keys). Audit logs and classified data have additional cryptographic integrity protection (SHA-512 hash chains).

Can I use my own encryption keys?

Yes. Customer-managed keys (CMK) via AWS KMS, Azure Key Vault, or HSM are supported. For air-gap deployments, customer-controlled key infrastructure is the default.

Does Trunnion AI log everything?

Yes. Every agent action, tool invocation, policy decision, user action, and system event is logged. Audit logs use SHA-512 hash chaining for tamper-evidence. Operational logs use standard structured logging patterns and integrate with customer SIEM systems.

Can I export Trunnion AI audit logs to my SIEM?

Yes. Audit and operational logs can be exported to standard SIEM platforms (Splunk, Elastic, Sentinel, Sumo) via syslog, REST APIs, or log forwarders. Integrity verification of hash-chained audit logs continues regardless of SIEM destination.

How does Trunnion AI handle observability?

OpenTelemetry-based instrumentation provides distributed tracing, metrics, and structured logs. Standard observability backends (Datadog, New Relic, Prometheus + Grafana, Honeycomb) are supported. Per-agent and per-workflow performance metrics are available out of the box.

Does Trunnion AI prevent prompt injection?

Trunnion AI mitigates prompt injection through multiple layers: input sanitization, structured tool call schemas (vs free-form text), classification enforcement on agent inputs and outputs, ABAC policy evaluation before tool execution, and HITL gates on high-consequence actions. No system is fully immune; the architecture limits damage when injection succeeds.

How does Trunnion AI handle hallucinations?

Trunnion AI mitigates LLM hallucinations through retrieval-augmented generation (grounding agents in verified data), tool-based validation (agents call APIs to verify claims), structured output schemas (constraining what agents can produce), HITL approval on high-consequence outputs, and continuous monitoring for output anomalies.

Can agents access the internet?

Internet access is configurable per agent and per deployment. Cloud deployments may grant agents specific external API access (allow-listed). Air-gap deployments have no external connectivity. ABAC policy controls which external resources each agent can reach.

What happens when an agent fails?

Agent failures are caught, logged, and handled per workflow configuration: retry with backoff, route to a fallback agent, escalate to a human via HITL, or fail the workflow. State is durably persisted so retries resume from the last checkpoint, not the start.

Can I version-control Trunnion AI workflows?

Yes. Workflow definitions are declarative and version-controlled. Customers manage workflow definitions in Git repositories, review changes via PR workflows, and deploy through CI/CD. Workflow versioning at runtime ensures in-flight executions complete on the version they started with.

How do I monitor agent quality over time?

Trunnion AI provides per-workflow and per-agent quality metrics: success rate, HITL rejection rate, escalation rate, output validation pass rate, and latency. Trends over time reveal workflow drift. Custom quality scoring can be added per workflow as needed.