Research Preview

AgenticBox

Your AI crew, your hardware, your data.

Replace routine business roles with AI agents that run entirely on your hardware. Zero cloud costs. Complete data privacy. Built on the Captain-Crew philosophy — you remain in command while AI agents handle execution, earning autonomy through performance and trust.

At a Glance

100%
Local Processing
6
Roles Automated
Zero
Recurring Costs
Open
Source Stack

Your AI Workforce, On Your Terms

AI assistants are powerful but cloud-dependent and expensive. API costs scale with usage, your data leaves your premises with every query, and generic AI doesn't understand your business context. AgenticBox changes the equation entirely.

The Problem

Cloud AI means recurring costs that scale unpredictably. Every query sends your data to third-party servers. Generic models lack context about your operations, your clients, and your industry.

The Solution

AI agents that live on your hardware, learn from your documents, and cost nothing to run after initial setup. Your data never leaves your premises. Your AI understands your business because you teach it.

The Philosophy

We call it Captain-Crew. AI agents are crew members — capable, tireless, always available. You're the captain — strategic decisions, quality control, final authority. AI earns autonomy progressively, through demonstrated performance and your feedback.

Six Roles, One Platform

AgenticBox automates routine tasks across six common business functions. Each AI agent specialises in a specific domain, working collaboratively under human oversight.

Network Engineer Agent

Monitor network health across sites, detect and diagnose connectivity issues, generate performance reports, and recommend optimisations. Always watching, always reporting.

Systems Administrator Agent

Monitor server health — CPU, RAM, disk, uptime. Automated backup verification, patch management recommendations, and common incident response. Your infrastructure, constantly supervised.

Graphic Designer Agent

Generate marketing visuals from templates, resize and optimise images for web, create social media graphics. Brand-consistent outputs without the design bottleneck.

Web Developer Agent

WordPress maintenance automation, plugin and theme update management, performance optimisation, and security hardening recommendations. Your web presence, automatically maintained.

Web Graphics Designer Agent

Landing page creation from briefs, A/B testing variations, image optimisation pipelines, and responsive design automation. From concept to deployment, accelerated.

Office Administrator Agent

Email triage and categorisation, invoice processing and tracking, proposal generation from templates, and calendar management assistance. Your back office, always running.

Agents augment your team — they don't eliminate it. Human oversight remains essential for quality and strategic decisions.

Transform Fixed Costs to One-Time Investment

The traditional model of cloud AI means recurring costs that grow with every query. Multiple SaaS subscriptions, per-seat licensing, and API charges that scale unpredictably.

Traditional Model

Recurring & Scaling Costs

Multiple part-time roles, cloud AI API charges per query, various SaaS subscriptions, and data leaving your premises with every interaction.

  • Ongoing salary costs for routine tasks
  • Per-query API charges that scale
  • Multiple SaaS subscription fees
  • Data sovereignty concerns
  • Unpredictable monthly costs

AgenticBox Model

One-Time Investment

A single hardware investment. After setup, your AI agents run at zero marginal cost. No API fees, no cloud subscriptions, no data leaving your infrastructure.

  • One-time hardware investment
  • Zero ongoing API costs
  • Zero cloud subscriptions
  • Complete data sovereignty
  • Predictable, fixed infrastructure costs

Actual savings depend on your current costs and automation scope. We help you model the ROI for your specific situation.

Your Data Never Leaves

Every AI interaction, every document indexed, every workflow executed — it all happens on your hardware. No exceptions.

Complete Local Processing

All AI inference runs on your hardware. No cloud APIs, no data transmitted to third parties, no external providers training on your information. Your conversations with AI stay in your building.

GDPR by Design

When data never leaves your premises, compliance becomes straightforward. No data processing agreements with AI providers, no cross-border data transfers, no third-party processor risk.

Air-Gap Capable

After initial setup and model download, AgenticBox can operate without internet connectivity. Perfect for sensitive environments, classified operations, or organisations with strict network policies.

Open-Source AI Stack

Built entirely on proven open-source components. No proprietary lock-in, no vendor dependency, full transparency.

Ollama

LLM Inference

ChromaDB

Vector Search

n8n

Workflows

Open WebUI

Interface

Docker

Deployment

Multiple Models

Run Llama, Mistral, CodeStral, Qwen, DeepSeek, and dozens more. Hot-swap models without downtime. Choose the right model for each task — coding, writing, analysis, conversation.

RAG Knowledge Base

Index your documents, proposals, procedures, and policies into ChromaDB. AI answers grounded in your actual business context — not generic internet knowledge.

Visual Workflows

n8n provides no-code workflow building connecting AI agents to your business systems. Email processing, notifications, integrations with hundreds of services.

Runs on Consumer Hardware

No data centre required. AgenticBox runs on high-performance consumer hardware, keeping the investment accessible.

Recommended

Hardware Specification

A single high-performance workstation handles all AI agents, model inference, and workflow orchestration.

  • Modern multi-core processor (16+ cores)
  • 96GB RAM for large model support
  • 1TB NVMe for OS and models
  • 4TB+ storage for business data
  • Fast SSD essential for model loading

Connectivity

Network Requirements

High-speed internet required for initial setup and model downloads. After deployment, AgenticBox can operate entirely offline.

  • Internet for initial setup only
  • Air-gapped operation supported
  • Optional secure remote access
  • No ongoing cloud connectivity required
  • We advise on specific configurations

Research Preview — Active Development

AgenticBox is in active development. We're transparent about where we are and where we're headed.

1

Foundation

Core infrastructure, AI inference, RAG system, and network architecture. Currently operational.

2

Automation

Email automation, document processing, monitoring dashboards. Currently expanding capabilities.

3

Autonomy

Progressive autonomy with feedback loops. AI agents earn trust through demonstrated performance.

4

Advanced

Multimodal capabilities, advanced automation, full-spectrum business agent coverage.

What's Working

Hardware infrastructure (100%), AI inference engine with multiple models (80%), RAG system with document indexing (100%), network and security architecture (100%). The foundation is solid.

What's Next

Email automation (expanding), monitoring dashboards (planned), progressive autonomy system (in design), advanced multimodal capabilities (roadmap). Target: full deployment within 12 months.

Be an Early Adopter

We're looking for forward-thinking organisations to join our pilot program and help shape the future of local AI automation.

Who We're Looking For

Organisations with 5–50 employees, currently spending significantly on routine IT and administrative tasks. Concerned about AI data privacy. Open to collaborative development and willing to provide regular feedback.

What You Get

Early access to AgenticBox, custom configuration for your use case, direct input into feature development, priority support during the pilot, and preferential terms for production deployment.

What's Required

Hardware investment (we advise on specifications), staff time for configuration and feedback, six-month minimum pilot duration, and regular feedback sessions with the IWH team.

We Practice What We Preach

IWH runs local AI infrastructure for its own operations. We bring deep expertise in IT infrastructure, security, and compliance to every deployment.

Our Experience

Built on Real-World Use

We run AgenticBox internally. Every feature has been tested in production against real business requirements before being offered to clients.

  • Running local AI for our own operations
  • Deep IT infrastructure expertise
  • ISO 27001, GDPR, NIS2 knowledge
  • Integration with real business systems
  • Complex deployment track record

What We Deliver

Full-Service Deployment

From hardware recommendation through to training your team, we handle the complete lifecycle.

  • Hardware recommendation and procurement
  • Full deployment and configuration
  • Document indexing and RAG setup
  • Custom workflow creation
  • Team training and ongoing support

Common Questions

Is this production-ready?

Not yet. This is a research preview. We're running it internally and seeking pilot partners to refine the platform before general availability.

What models can it run?

Any Ollama-compatible model — Llama 3.3, Mistral, CodeStral, Qwen, DeepSeek, and many more. Model selection depends on your hardware and use case.

Do I need technical staff?

Basic operation requires minimal technical knowledge. Configuration and customisation benefit from IT experience, which IWH provides during setup and ongoing support.

What about model updates?

New models are released regularly. Updating is a straightforward process, and we can handle this as part of ongoing support arrangements.

Can it integrate with existing systems?

Yes. n8n provides integrations with hundreds of services. Custom integrations can be developed for your specific systems and workflows.

What happens to my data?

Your data stays on your hardware. Period. No cloud transmission, no external processing, no training data sharing. Complete data sovereignty.

Investment Model

AgenticBox follows a one-time investment model. Hardware costs depend on your requirements, and IWH provides deployment, configuration, and ongoing support services.

  • Hardware investment (we advise on specifications)
  • Deployment and configuration services
  • Document indexing and RAG setup
  • Optional ongoing support agreement
  • No per-query or per-user recurring fees

Pilot program participants receive preferential terms. Contact us to discuss your requirements.

Discuss Requirements

Ready to explore local AI?

AgenticBox is in active development. We're looking for forward-thinking organisations to join our pilot program and shape the future of local AI workforce automation.

Apply for Pilot Program Back to All Products