Sentinel Alpha

Caprica Vault

Private AI inside infrastructure you control

Private AI workspace
NL

Private AI for work that cannot become public data.

Caprica Vault runs inside your own environment. Source files, access rules, and audit evidence stay under your control while a privacy firewall masks sensitive context before any approved external model call.

Self-hosted with Docker
Outbound-data preview before model calls
Local model or approved AI provider
Local by default

Keep source files, indexes, roles, and audit events in your environment

Policy checked

Detect and mask sensitive data before an external model step

Human approved

Review the outbound payload and generated result before handoff

Enter your email and we will follow up with a private demo path for the Caprica workflow that fits your team.

Visible model boundary

Policy checked

Your controlled environment

Caprica Vault

Source files
Vector index
Roles
Audit trail

Privacy firewall

Detect

Mask

Preview

Log

Model route

Local model

The complete model step stays inside.

Approved provider

Receives sanitized text snippets only.

Raw identifiers, full files, and audit events stay outside the external payload.

Caprica workspace

Source input

+

Add documents, notes, or interface specs

Release notes, process notes, contracts, mappings, and supporting files

Unit4 Release Notes Q2.pdf

Release document

Ready

Supplier Portal Mapping.xlsx

Integration mapping

Ready

Approval Process Notes.docx

Business process

Ready

Question asked

Which roles must validate this change before the purchase flow can go live?

Answer returned

  • Caprica identifies which roles, approvals, and business risks are affected.
  • It suggests a role-based validation flow with checkpoints and approval steps.
  • The result is ready to export for review, audit, or supplier follow-up.
Source: release notesRole routing includedData stays inside

Why teams buy this

Ask questions in plain business language.

Turn source material into usable drafts faster.

Keep privacy, review, and rollout under control.

FM

Pilot feedback

Within a day, we could ask questions we normally avoided because the data could not leave our environment.

Functional application manager, Dutch healthcare organizationTested in a 14-day Starter trial

How it works

From Docker image to reviewed result in five clear steps.

The system administrator controls the environment. Key users build trusted workspaces. Employees use simple workflows. Caprica guards the model boundary and records the evidence.

Caprica Vault

The controlled route

Your source data stays local; the model route is a deliberate choice.

Stays local

Files, index, roles

Privacy firewall

Detection, masking, outbound preview

Model choice

Local or sanitized provider call

Back in the Vault

Human review and audit evidence

Open the interactive privacy demo ->
  1. 01

    System admin

    Deploy and set the boundary

    Run the signed Docker image on a server, PC, or private cloud and configure roles, storage, network policy, and the allowed model route.

  2. 02

    Key user

    Create a trusted workspace

    Upload approved documents, release notes, API specifications, or process files. Caprica indexes them locally and applies workspace access.

  3. 03

    Business user

    Choose a practical workflow

    Ask a knowledge question, prepare release validation, design an integration, or run a controlled analysis without learning prompt engineering.

  4. 04

    Privacy firewall

    Inspect what the model may receive

    Sensitive values are detected and masked. With an external provider, the outbound preview shows the sanitized context; with a local model, no provider payload leaves.

  5. 05

    Reviewer

    Validate, approve, and export

    Check sources and results, record sign-off, and export the answer, test set, flowchart, or mapping together with an auditable history.

The problem

Most AI tools are not built for sensitive business work.

Teams want AI, but the default options send data to the cloud, require too much technical setup, or still leave people doing manual repeat work.

01

Public AI puts data in the wrong place

That is a real risk for finance, legal, healthcare, research, and any team working with confidential documents or internal process logic.

02

Private AI should not take months of IT work

Teams do not want to stitch together models, security, approvals, exports, and workflows before they can prove value. Caprica is the ready-to-run version.

03

Manual analysis still slows delivery

Document review, release validation, functional design, and mapping work still take too much time when people start from scratch every time.

The solution

One controlled workspace between your data, your people, and AI.

Caprica combines private retrieval, policy controls, specialist workflows, and human approval so smaller IT teams do not have to assemble an enterprise stack themselves.

1

Keep the source of truth inside

Documents, indexes, permissions, and audit evidence stay inside the infrastructure your organization controls.

2

Control what a model may see

The privacy firewall detects identifiers, masks sensitive context, and shows the outbound payload before an approved provider is used.

3

Turn answers into governed work

Run document Q&A, release validation, and integration design with human checkpoints, role-aware review, and exportable evidence.

Inside the Vault

Four practical workspaces, one private boundary

Start with one module and expand when the team is ready. Each workspace uses the same access controls, privacy firewall, and audit trail.

Knowledge Vault

Ask internal documents in plain language

Search contracts, policies, manuals, and project files through a private RAG workspace with source-aware answers.

Validation Hub

Turn release notes into role-based tests

Generate test questions, checkpoints, owners, and sign-off evidence for each affected business role.

Integration Hub

Draft integrations on top of existing APIs

Use API specifications and process notes to prepare flowcharts, mappings, business rules, test scenarios, and review packages.

Analysis Workspace

Create evidence-backed operational insight

Compare controlled source material, calculate exact local results where needed, and prepare reviewable management summaries.

Privacy

Private means knowing exactly what stays and what may leave.

Caprica makes the model boundary visible. A local model keeps the entire flow inside; an approved external provider receives only the sanitized context permitted by policy.

A privacy boundary your team can inspect

Deploy Caprica in Docker, set roles and network policy, and choose a local model or your own provider. Before an external call, users can inspect the masked outbound payload.

Source files and indexes stay local
PII detection and masking
Tamper-evident local audit trail
1

Local source, index, permissions, and evidence

Original files, vector data, role settings, and audit events remain inside infrastructure your team controls.

2

Preview the external model payload

Caprica shows the sanitized text eligible for an external call and keeps raw identifiers and source files outside that payload.

3

Choose the route per deployment

Use a local model for a closed route, or allow a configured provider after policy checks, masking, and human review.

Benefits

Less manual work. More confidence.

Caprica helps teams move faster without losing oversight.

Faster answers

Find, compare, and summarize information in minutes instead of manual file-by-file work

Better adoption

Give teams an AI workflow they can trust instead of another tool they need to work around

Less rework

Start from structured drafts for validation, mapping, and handoff instead of a blank page

FAQ

Common questions

Straight answers in plain language.

What kinds of files can Caprica work with?

Teams usually start with PDFs, Word documents, spreadsheets, release notes, mapping sheets, and process notes.

Who is Caprica for?

Caprica is built for teams that handle sensitive information and still want practical AI support. Typical buyers are in finance, legal, operations, QA, and functional management.

Can it stay inside our own environment?

Yes. That is one of the core reasons Caprica exists. You choose where it runs and how access is managed.

What is the difference between Starter, Pro, and Pro+?

Starter is for private document Q&A. Pro expands that into a private workspace for recurring business use. Pro+ adds Validation Hub and Integration Hub for testing and design workflows.

What AI provider do we need?

You can use a local model or bring an approved provider such as OpenAI, Claude, or Gemini. External calls receive only the policy-approved context shown in the outbound preview.

How long does setup take?

For a first pilot, most teams are running in under two hours with Docker and the step-by-step guide.

Can we start with one workflow and expand later?

Yes. That is the best way to evaluate Caprica. Start with one real use case, prove value, and then expand into other workflows.

Next step

One workflow. Your own environment. Fourteen days.

Show your team what private AI looks like when it is built for real business work, not just demos.

Enter your email and we will follow up with a private demo path for the Caprica workflow that fits your team.