Mobile Keep™
Portable case
Self-contained for a private office, branch location, field deployment or secure demonstration.
Sovereign Enterprise Intelligence Platform
Own the operating center of your business.
The Keep™ is a private, company-owned intelligence hub that unifies your data, workflows and AI under your control.
- 90 days
- from first conversation to a working hub
- 4 profiles
- portable, rack, private cloud, boardroom
- $17.5k–$75k
- one-time deployment, published upfront
Your team did not vote to move customer notes, pricing logic, strategy and institutional memory into a fragmented tool stack. It drifted there.
The question beneath every AI decision
Where does the intelligence of your company live?
The product
The stronghold at the center of your business.
One secure environment you own and control, holding data, customer knowledge, AI agents, workflows, documents and systems. Unlike a rented platform, it keeps your most valuable asset inside the company.
- Company knowledge
- Customer memory
- Documents & records
- AI conversations & outputs
- Workflows & automations
- Policies & governance
- Operational intelligence
- Institutional memory
Deployment profiles
Four ways to deploy.
Size for the workload, not the label. Every Keep includes secure boot, TPM-backed key storage, full-disk encryption, remote management, monitored backups and a documented replacement path.
Keep Rack™
1RU rack server
Rack-mounted infrastructure for an office, data center or colocation environment.
Keep Vault™
Private cloud
A dedicated deployment inside your own Virtual Private Cloud, with no onsite appliance.
Executive Keep™
Boardroom edition
A visible statement of strategic control, at the level where direction is set.
Deployed and supported through your MSP or authorized reseller. Find a partner
Reference architecture
Everything flows in. Nothing leaks out.
Approved sources feed a hub you control. Governed AI, workflows, memory and communications run inside your boundary — and the outputs return to your people, not to a vendor's platform.
01 Input Approved sources you already own
- CRM Accounts, pipeline
- Email Threads, intent
- Documents Contracts, decks
- Files Drives, archives
- Web Approved sources
- APIs Line-of-business
02 Transformation Governed work, inside the boundary
The Keep™ Customer-controlled boundary
- 01 Agents
- Local and approved cloud AI acting on your behalf
- 02 Workflows
- Automation that removes repetitive work
- 03 Knowledge
- Private memory, embeddings and retrieval
- 04 Comms
- Audience and customer engagement you own
- Dashboards The operating view
- Reports Board-ready
- Actions Work executed
- Notifications Routed to owners
- Exports Portable, yours
03 Output Returned to your people, not a vendor's platform
Every result returns as governed input — the loop compounds inside your boundary
Data, permissions, memory and workflow logic never leave the boundary. Approved external services connect through governed interfaces.
The danger is not that businesses use too many tools. It is that the tools are becoming the business.
Software stack
Open-source-first, not open-source-only.
The objective is control, interoperability and operating value — not ideological purity. Components are chosen by function and replaced as technology, licensing or your needs change.
| Component | Role | What it does | License posture |
|---|---|---|---|
| Ollama | Local AI runtime | Runs supported models on customer-controlled hardware. | MIT runtime; per-model licenses |
| Open WebUI | Private AI interface | Self-hosted access to models, knowledge and assistants. | Branding provisions — review |
| PostgreSQL | Structured data | Operational, application and customer data foundation. | PostgreSQL License |
| pgvector | Semantic memory | Vector storage and similarity search for embeddings and RAG. | PostgreSQL-style |
| Crawl4AI | Governed ingestion | Collects approved web content for AI-ready knowledge. | Apache 2.0; attribution |
| Whisper | Speech-to-text | Turns meetings and media into searchable text, locally. | MIT code and weights |
| n8n | Workflow orchestration | Connects models, data, email, documents, CRMs and APIs. | Fair-code Sustainable Use |
| WordPress + FluentCRM | Owned publishing | Content, portals, campaigns and customer journeys. | GPLv2+; Pro tiers commercial |
| EspoCRM | Self-hosted CRM | Accounts, contacts, opportunities and customer history. | AGPLv3 — review obligations |
License summaries are planning guidance only. Confirm current source, dependencies, redistribution, branding and commercial terms before bundling or committing to a deployment.
Physical configurations
Three reference profiles.
Planning profiles, not a frozen bill of materials. Final sizing depends on model choice, concurrency, data volume, availability requirements and your partner's security standard.
| Specification | Small | Medium | Large |
|---|---|---|---|
| Typical users | 25–100 | 100–500 | 500–2,500+ |
| Workflow scale | 1–3 priority | 5–15 production | 15–50+ and agent services |
| CPU | 12–16 cores | 24–32 cores | 32–64 server-class |
| Memory | 64 GB ECC | 128 GB ECC | 256–512 GB ECC |
| AI acceleration | Optional 16–24 GB | 32–48 GB VRAM | Dual 48 GB or 80 GB-class |
| Usable NVMe | 2 TB mirrored | 4–8 TB protected | 8–32 TB expandable |
| Networking | 2.5–10 GbE | Dual 10 GbE | Dual 10/25 GbE, segmented |
| Availability | Single node, encrypted backup | Redundant storage, spare unit | HA, redundant power, failover |
Local model performance varies materially by architecture, quantization, prompt length and concurrent demand. Benchmark the actual workload before procurement.
Live reference deployment
Proof in production: ChargedUp.work.
A spin-out from PublioSTUDIO, built in 90 days on sovereign architecture. The team stopped assembling the machine every week and started using it.
20 6 hours per week to publish and distribute
90 days from first build to a working hub
300+ published articles in owned memory
3 days to onboard new operators to production
The architecture that produced this result is the architecture on offer.
Transparent pricing
You see the math before you commit.
The decision to own is easy at $17,500. It becomes prohibitive at an unexpected $750,000.
Deployment · one-time, 90 days
$17,500–$75,000
- Discovery and workflow mapping
- Server provisioning and security baseline
- Local AI model and RAG training on your content
- Self-hosted CRM and customer data layer
- Three to five priority workflows
- Team training and runbook
Platform fee · monthly
$700–$3,500/mo
- Managed intelligence layer (tuning, patterns)
- Monitoring, backups, security patching
- Platform improvements and new integrations
- Deployment-aware AI support assistant
- Customer roundtables and skills transfer
- Quarterly strategy reviews
Services such as AI models, middleware and email delivery are pass-through and outlined in the proposal.
The low-risk path
From conversation to working hub in 90 days.
A governed micro-deployment with named outcomes, not a platform mandate. Every phase has an exit gate — the work does not advance until it is done, and it does not drift either.
- Day 0
Sovereignty conversation
Thirty minutes. Map the stack and find the right first move.
- Week 1
Kickoff and intake
Confirm goals, scope, RACI, access and success metrics.
- Weeks 2–4
Pilot deployment
Core hub, integrations and first workflows live in QA.
- Weeks 5–8
Operating cadence
Team trained. Runbook approved. Real outputs flowing.
- Day 90
Review and expand
Measured outcomes. Expansion backlog. Managed operations.