Strategic Vision Document

Gloo AI
Agent Strategy

How Gloo AI becomes the agent infrastructure for values-aligned organizations — enabling every team, partner, and developer to build, deploy, and govern autonomous agents on a shared, trusted platform.

Gloo AI · February 2026 · Confidential

Executive Summary

The AI industry is shifting from model intelligence to agent infrastructure. The companies that win will not be the ones that build the most agents — they will be the ones that build the platform every agent depends on.

Gloo AI already operates the most comprehensive values-aligned AI infrastructure in the faith and flourishing ecosystem. The Flourishing Engine wraps frontier models with safety, routing, and alignment. The Data Engine transforms content into searchable, rights-managed, AI-ready knowledge. Studio gives developers and teams the tools to build on top.

This document presents the strategy for the third foundational engine: the Agent Engine. It is the orchestration and governance layer that enables any team — internal or external — to build, deploy, monitor, and govern autonomous agents on shared infrastructure. It completes the three-engine architecture and positions Gloo AI as the definitive agent platform for our ecosystem.

The strategic position: Gloo AI owns the agent infrastructure — runtime, deployment, memory, tooling, monitoring, and governance. Application teams like MinistryChat build interfaces and agent marketplaces, extending their current workflows and skills. Service teams like Gloo 360 deliver white-glove agent deployments for enterprise customers. Everyone builds on the same trusted foundation.

This mirrors the approach the largest AI companies are taking. OpenAI launched Frontier in February 2026 — an enterprise agent management platform with shared business context, execution, evaluation, and governance. OpenClaw, the open-source agent with 145k GitHub stars, demonstrated that when you provide the runtime, memory, tools, and channels, an unlimited number of use cases emerge on top. Both validate the infrastructure-first thesis. Neither provides what Gloo uniquely can: values alignment, theological accuracy, and the ontology of faith-based org data, operational data, and intellectual property.

The Agent Evolution

Our teams have been building real, valuable AI capabilities — chat assistants, automated workflows, smart prompt systems, and domain-specific skills. These solve genuine needs in the marketplace. The agent infrastructure within Gloo AI will extend these capabilities and enable more offerings for the services and application teams.

Where We Are Today

Across Gloo, teams have shipped AI capabilities that deliver real value. Each represents a stage on the path toward full agent capability — and each is a necessary building block.

Foundation
Chat Assistants

Conversational interfaces that answer questions and help users think through problems. These are valuable and will remain a core interaction pattern. What they lack is action-oriented agency — they are constrained to independent conversational use cases.

Foundation
Automated Workflows

Predetermined sequences of steps that automate manual processes. These drive efficiency gains for well-defined, repeatable tasks. The limitation is that workflows must be explicitly authored for every scenario — even embedding a reasoning model as a step still constrains the system to a finite set of predefined paths rather than adapting fluidly to novel situations.

Foundation
Smart Prompt Systems

Applications that construct sophisticated prompts on behalf of users — making AI easier to use by encoding domain expertise into the prompt layer. This is good UX engineering. This was the foundational step to AI capabilities in SaaS.

Foundation
Domain-Specific Skills

Targeted capabilities within a chat interface — sermon prep, content analysis, summarization. These encode valuable domain expertise. When you add persistent memory, tool access, and the ability to operate autonomously toward a goal, a skill evolves into an agent.

Where We're Going

Autonomous Agents: Where the Biggest Opportunity Lives

A Gloo agent is a long-lived, goal-oriented entity that combines four components — a Brain (reasoning model), a Harness (guardrails and governance), Memory (persistent context across sessions), and Tooling (data access and actions) — to take autonomous action and produce outcomes within an aligned, safe environment. It persists across sessions, accumulates context, selects its own tools, executes autonomously for hours or days, and operates within enforced safety boundaries. Think of a research analyst that works overnight and hands you a finished deliverable in the morning — not a chatbot that resets every conversation.

The market is moving here fast. Gartner called agent management platforms "the most valuable real estate in AI." OpenAI, Anthropic, Google, Salesforce, and a wave of startups are all building toward agents. Now Gloo has the opportunity to provide the agent infrastructure for our ecosystem that has the values alignment and domain-specific capabilities our market needs.

Figure 1 — Anatomy of a Gloo Agent
GLOO AGENT Brain Model-agnostic reasoning Any frontier model — OpenAI, Anthropic, Google, open-source — can serve as the brain. Intelligent routing selects by cost, capability, and latency per task. Harness Guardrails & governance System-level instructions and constraints. Flourishing Engine safety layers enforced by default. Values alignment on every call, regardless of who built the agent. Memory Long-term continuity Short-term session context. Long-term user facts with explicit consent. Shared team context with access controls. This is what makes agents long-lived. Tools Context & actions Data Engine retrieval (RAG), API calls, workflow execution, external system integration. Domain-specific tool registry shared across the ecosystem.

Market Context

Two developments in the last week validate the infrastructure-first approach and offer architectural patterns worth studying closely.

OpenAI Frontier

On February 5, 2026, OpenAI launched Frontier — an end-to-end enterprise platform for building, deploying, and managing AI agents. Frontier treats agents as "AI coworkers" and organizes around four pillars: shared business context, agent execution, evaluation and optimization, and security and governance.

Figure 2 — OpenAI Frontier Architecture (Reference)
OpenAI Frontier Enterprise security and governance designed for sensitive and regulated work Interfaces ChatGPT Enterprise ChatGPT Atlas Business Applications Agents Your Agents OpenAI Agents Third-Party Agents OpenAI Frontier Evaluation and Optimization Built-in evaluation and optimization loops to improve performance Agent Execution Model intelligence and tools for agents to plan, act, and recover real-world tasks Business Context Shared business context across data, systems, and workflows

Early adopters include Uber, State Farm, Intuit, and Thermo Fisher Scientific. Gartner called agent management platforms "the most valuable real estate in AI." This is the category we need to own for values-aligned organizations.

OpenClaw

OpenClaw (formerly Clawdbot) is the open-source agent platform that went viral in early 2026, collecting 145k+ GitHub stars. Its architecture tells a different but equally instructive story: a hub-and-spoke gateway that serves as the central control plane. Channels (WhatsApp, Telegram, Chrome, Slack) feed into the gateway, which handles agent runtime, session memory, tool routing, and security policies. Everything else — every professional service, every interface, every integration, every use case — builds off of the infrastructure.

This is how a single open-source project spawned thousands of use cases: by providing infrastructure and letting the ecosystem build on top. The lesson is clear. The governance gap, however, is also clear — Palo Alto Networks flagged OpenClaw's combination of private data access, untrusted content exposure, and external communication ability as a serious enterprise risk. This is one of the critical gaps Gloo's AI infrastructure was built to fill: trust.

Figure 2B — OpenClaw Hub-and-Spoke Architecture
Channels WhatsApp Telegram Chrome Slack GATEWAY Message Router Session Mgmt Auth & Routing Security Policies Rate Limiting AGENT RUNTIME Orchestration Planning & Reasoning Tool Calling Context Assembly Response Generation MCP Protocol Layer LLM Model Inference Memory Short-term · Long-term Conversation State Skills Modular Capabilities Web Search File Operations Code Execution Custom Tools COMPUTE RESOURCES

What Both Validate

Frontier validates that enterprises need a unified platform for business context, execution, evaluation, and governance. OpenClaw validates that when you provide the runtime, memory, tools, and channels, an unlimited number of use cases emerge.

Where Value Accrues

The organizations that own the full stack — data, trust, infrastructure, and execution — build the strongest moat. When Gloo handles this entire stack, we become the most valuable entity in our market.

In the AI landscape, value concentrates with whoever controls the layers that are hardest to replicate. Individual agents can be copied. Workflows can be rebuilt. But proprietary data, institutional trust, and a purpose-built control plane compound over time and become increasingly difficult for competitors to match. Every layer of the stack Gloo owns reinforces the others.

Proprietary Data & Content
Longitudinal history, organizational context, operational exhaust — compounds over time
Trust, Governance & Control Plane
Identity, safety, audit, compliance, values alignment — the layer enterprises require
Infrastructure & Compute
Model routing, inference, tool orchestration, memory management
Agent Execution
Runtime, deployment, session management, observability

Gloo owns every layer. Gloo AI provides the infrastructure — data, trust, compute, and execution. Gloo 360 brings managed agent deployments for enterprise organizations, operating and optimizing agents on behalf of clients. MinistryChat provides the interface where agents are deployed, observed, and interacted with at scale. When the full stack is owned across infrastructure, services, and application — and each layer reinforces the others — the result is a compounding moat that deepens with every agent deployed, every dataset ingested, and every organization onboarded.

Three-Engine Architecture

Three foundational engines. Each solves a distinct hard problem. Together, they provide everything an agent needs to operate reliably, safely, and at scale.

Figure 3 — Gloo AI Three-Engine Architecture
SERVICES & APPLICATIONS MinistryChat Gloo 360 Capital Partners Developers Internal Ops AGENT ENGINE Orchestration & Governance Agent Runtime Config → Execute Tool Registry Shared tools Sessions Chat · Async · Stream Memory Short · Long · Shared Deployment Version · Rollback Observability Logs · Cost · Alerts · Safety Governance — Cost Limits · Approval Gates · RBAC · Audit Logs · Compliance · Values Enforcement FLOURISHING ENGINE Model Harness Input Safety Sentinel Screening Intelligent Routing Core · Core Select · Select Contextual Fabric Context Engineering Integrity Layering Constitutional AI DATA ENGINE Content Intelligence Ingestion Docs · Media · Web AI Enrichment 95+ metadata fields Vector Storage Weaviate · Typesense Semantic Search RAG Retrieval Agent Engine — orchestration, lifecycle, governance Flourishing Engine — safety, routing, alignment Data Engine — content, RAG, grounding

How Each Engine Serves Agents

Engine Function What It Provides to Agents
Agent Engine Orchestration & Governance Runtime execution, deployment (version control, rollback), session management (chat, async, streaming), memory (short/long/shared), tool registry, observability, and governance (RBAC, audit, approval gates)
Flourishing Engine Model Harness & Safety Every LLM call flows through the Flourishing Engine regardless of who built the agent. Input safety, intelligent routing, contextual fabric, model inference, and integrity layering. Six dimensions of safety enforced by default.
Data Engine Content Intelligence Rights-managed, organization-scoped datasets. Content ingestion, AI-powered enrichment, vector storage, and semantic retrieval. The grounded context agents need to do real work.

How It Works Together

The agent infrastructure creates two distinct go-to-market motions on a single platform. One is high-touch and tailored for enterprise. The other is self-serve and designed for scale. Both depend on the same foundational engine.

This is the same pattern the most successful platforms in tech follow — AWS provides infrastructure, and both Accenture (services) and Vercel (self-serve) build on top of it for different buyer profiles. The infrastructure doesn't care whether the agent was configured by an enterprise solutions team or by a developer in a web interface. It provides the same runtime, safety, grounding, memory, and governance regardless.

Figure 5 — The Gloo AI Agent Delivery Model
Enterprise Agent Services Gloo 360 Tailored agent solutions built and operated for enterprise organizations. Dedicated engineering teams design, deploy, and continuously manage agents alongside the customer. High-Touch Custom SOWs Managed Self-Serve Agent Platform MinistryChat & Developer Ecosystem Agent marketplace where teams and developers discover, configure, and deploy agents on their own. Built for speed and accessibility — no engineering engagement required. Self-Serve Marketplace Usage/Subscription-Based GLOO AI AGENT ENGINE Shared Infrastructure Agent Runtime Tool Registry Memory Sessions Deployment Observability Metering & Billing Flourishing Engine Data Engine Governance & RBAC Values Alignment Safety Layers RAG & Search

The infrastructure is indifferent to the go-to-market motion above it. Whether an agent is hand-crafted by a Gloo 360 solutions architect for a Fortune 500 customer or configured in five minutes through a MinistryChat template by a church administrator, it runs on the same runtime, benefits from the same safety guarantees, and is governed by the same compliance infrastructure. That's the power of the platform model — invest once in infrastructure, enable unlimited distribution paths on top.

Autonomous Agents in Practice

These aren't chatbots that answer a question and reset. They aren't workflows that follow a predetermined sequence. These are persistent, goal-oriented agents that operate autonomously over days, weeks, and months — accumulating context, making decisions, and producing outcomes within safe, governed boundaries.

GLOO 360
Church Planting Agent

A denomination deploys a managed agent to accelerate its church planting strategy. The agent continuously analyzes demographic data, community need indicators, and existing church coverage to identify the highest-opportunity locations. It researches local partnerships, drafts outreach plans for prospective planters, and tracks each plant's progress from initial assessment through launch. Regional directors review recommendations and approve next steps at each stage. The agent keeps dozens of planting initiatives moving forward simultaneously — leadership focuses on vision and relationships.

MINISTRYCHAT
Church Operations Agent

A church administrator subscribes to a persistent operations agent that manages volunteer coordination, sends context-aware reminders, tracks attendance trends, and prepares a weekly staff briefing. The agent learns the church's rhythms over time — anticipating seasonal surges and flagging conflicts. Every Monday, the administrator reviews a summary and approves the week's plan. The agent handles execution. Staff stay focused on people, not logistics.

3RD-PARTY DEVELOPER
Bible Translation Project Agent

A major Bible translation organization builds on Gloo's agent infrastructure to deploy a translation project management agent. It tracks progress across dozens of language teams, surfaces bottlenecks, coordinates reviewer assignments, and flags passages that need theological review. Translation directors approve key decisions and quality checkpoints. The agent keeps a multi-year, multi-language project moving without letting any workstream stall.

3RD-PARTY DEVELOPER
Parachurch Donor Stewardship Agent

A national parachurch ministry builds a donor stewardship agent on Gloo's platform. It monitors giving patterns, identifies at-risk donors, drafts personalized outreach for the development team, and tracks engagement across events and campaigns. The development director reviews recommendations before any outreach goes out. The agent manages the pipeline continuously — the team focuses on relationships, not spreadsheets.

What makes these autonomous agents — not chatbots or workflows. Each persists over weeks and months, accumulating context and pursuing goals. Each adapts its approach based on what it learns. And in every case, humans remain in the loop at the decision points that matter — approving outreach, reviewing recommendations, setting policy, and making judgment calls. The agent handles the sustained effort. People stay in control of the outcomes.

Business Model

Three distinct monetization channels — each tied to a different go-to-market motion and buyer profile.

Channel Value Proposition Revenue Model Example
Gloo 360 Fully managed agent deployments — we build, configure, and operate custom agents on your behalf Implementation + recurring managed-service fees A denomination deploys a swarm of managed staff-support agents with 360 handling setup, training, and ongoing optimization
MinistryChat Persistent AI employees — long-lived agents that learn your organization, take autonomous action, and compound in value over time Premium subscription tiers priced on agent capability and usage A church administrator subscribes to an operations agent that handles volunteer coordination, follow-ups, and weekly reporting on its own
Gloo AI API Agent infrastructure as a service — build on our runtime, safety, and data layers without rebuilding the stack Pay-per-use on tokens, API calls, memory, and tool access A faith-tech startup builds a counseling-support agent on Gloo infrastructure, paying only for what it consumes

The Flywheel

More teams deploy agents, the tool registry grows. More shared tools, new agents are easier to build. More agents operating, Data Engine accumulates richer organizational context. Richer context, agents become more effective. Each cycle increases the value of the platform for every participant.

Figure 6 — Platform Flywheel
Gloo AI Platform More Teams Deploy Tools Grow Easier to Build Data Compounds Agents Improve Platform Value ↑

Implementation Roadmap

Learn the needs of the market through enterprise deployments upmarket. Over time, enhance the platform to enable higher-volume agent deployment downmarket. Enterprise developer clients extend our reach across the broader ecosystem.

Gloo 360 — Learn the Market Upmarket

Enterprise deployments through 360 are where we learn what the market actually needs — how organizations want to configure agents, govern them, and measure their impact. These high-touch engagements surface the real requirements that shape the platform's capabilities.

MinistryChat — Scale It Downmarket

As the platform matures from enterprise learnings, MinistryChat becomes the channel for higher-volume agent deployment. Capabilities proven through 360 engagements get productized into self-serve experiences — persistent agents, marketplace distribution, and subscription-based access for churches and ministries at scale.

YouVersion — Enterprise Developer Client

YouVersion represents the first enterprise developer client building on Gloo's agent infrastructure. Working alongside the largest faith-based app in the world teaches us what developer-facing platform capabilities need to look like — and extends our reach into a corner of the market we couldn't access through 360 or MinistryChat alone.

Magisterium — Enterprise Developer Client

Magisterium represents a second enterprise developer client, extending our reach into the Catholic ecosystem. Their requirements for doctrinal precision and institutional oversight shape how the platform handles content governance — capabilities that translate to any tradition with strict trust requirements.

Why this sequence matters. Enterprise deployments through 360 teach us what the market needs. MinistryChat productizes those learnings for volume. YouVersion and Magisterium prove the platform works for external developer clients across different segments. Each channel feeds the others — 360 surfaces requirements, MinistryChat validates scale, and developer clients validate extensibility.

Why This Is Right

Three alternatives exist. Each fails a different test.

Centralized development

The AI team builds every agent for every team. Doesn't scale. Platform team becomes a bottleneck. Teams wait months. Innovation slows to the speed of the smallest team.

Decentralized without infrastructure

Every enterprise services team builds agent capabilities independently. Fragmentation. Each team rebuilds runtime, deployment, monitoring, and governance from scratch. No shared learning. No consistency. No platform revenue opportunity.

Application-level agent features

Agent capabilities get embedded directly into individual applications as features. Produces standalone prompts and workflows, not long-lived agents. Missed platform opportunity. Governance inconsistency across the ecosystem.

Decentralized development on centralized infrastructure. Teams closest to problems build the solutions. The platform provides a unified control plane, enables network effects, opens new revenue, enforces governance by default, and establishes competitive differentiation no one else in our market can match. We're already building agents. The question is whether we capture the patterns in shared infrastructure or rebuild them independently every time.

Reference Sources

External

OpenAI Frontier — Enterprise AI agent management platform, launched February 5, 2026. Architecture: Business Context, Agent Execution, Evaluation & Optimization, Security & Governance. Early adopters: Uber, State Farm, Intuit, Thermo Fisher Scientific.

OpenClaw — Open-source agent platform (formerly Clawdbot). Hub-and-spoke gateway, channels, memory, tool routing, MCP integration. 145k+ GitHub stars.

Internal

Agent Engine: Platform Infrastructure for Decentralized Agent Development · Agentic Architecture for Gloo · Gloo AI: What It Is & How It Works · Gloo360 Reference Architecture · The Agent Economy Is a Mirage · Gloo AI Deck (Feb 2026)