Skip to content

Category: CoreAI · Private AI Operations

AI for work that cannot leak.

CoreAI is a bilingual private AI operations platform. Every prompt is classified for sensitivity and routed accordingly — sensitive content stays on local Apple Silicon inference, public-grade requests reach cloud reasoning through one audited gateway. Native iOS and macOS clients, offline-first.

Built for

  • Operating teams
  • Invited partners
  • Research collaborators
Status
Live
Ledger step
Ledger step 4 of 4
Family
Workspaces
Access
Invite only
Listed
4 / 4
Invite-only access
Production
Ember Edge inference
Local-first
Ember AI family
7 models
Native clients
iOS + macOS

What this page can show today

  • The register marks this track Live.
  • Capabilities below are the published feature list for this track.

Limits of this page

  • Access is invite-only. A public URL here is not an open signup.
  • This page does not publish financials, headcount, or client names.

Key features

  1. 01

    Private LLM gateway

    A single internal gateway brokers every model call across our platforms, with provider keys, rate limits and audit in one place.

  2. 02

    Sensitivity-aware routing

    Each prompt is classified and routed; sensitive content stays on local inference, public-grade requests reach cloud reasoning.

  3. 03

    Local-first inference

    Ember Edge, our on-device runtime for Apple Silicon, powers offline-first inference paths inside our products. Data stays on-device by default.

  4. 04

    Native iOS + macOS clients

    First-party SwiftUI clients backed by the private gateway; tools wired through structured, auditable tool calls for verifiable actions.

Published links

Technology stack

Swift · SwiftUI · Python · Ember Edge · Ember Gateway

Related platforms

Family on the register: Workspaces Same family first, then tracks at the same lifecycle stage.

  • Same family · Beta

    Apertura

    macOS Productivity

    This site only

  • Same family · Beta

    Aegis

    Cybersecurity & Trust Intelligence

    This site only

  • Same family · Beta

    Tether

    Session Continuity

    This site only

  • Same stage · Live

    CarbonBridge

    Carbon Compliance · CBAM

    Public product site

This cell on the investor register

Status and family write into the /investors URL. The holdings table lists every public track in that cell.

Open Live Workspaces holdings
4 / 4

JSON fields are copied from this row in src/data/platforms.ts. No financials.

How this track works

Audience on the left, the published Ember stack in the middle, capabilities on the right. Every node is copied from this track in the public register.

How this track moves work from its published audience through Coreidea and Ember into listed capabilities.

Inputs

Coreidea and Ember

Outputs

Select a node to inspect the published role and capability.

Convenient AI and confidential work pull in opposite directions.

Teams paste contracts, financials and strategy into consumer chatbots every day — because the sanctioned alternative is usually nothing at all.

  • Shadow AI is the default

    Without a sanctioned tool, sensitive material flows to whichever public chatbot each employee prefers — unlogged, unredacted, unrecoverable.

  • One-size routing fits nobody

    Blanket bans kill productivity; blanket cloud access leaks data. The real requirement is per-prompt judgment, applied consistently.

  • No audit trail, no accountability

    When something goes wrong, most teams cannot answer the basic questions: what was sent, to which model, by whom, and what came back.

Policy lives in the gateway, not in goodwill.

CoreAI does not ask people to remember the rules. The platform classifies, redacts, routes and logs every call — automatically, the same way, every time.

  • Sensitivity-aware routing

    Each prompt is classified before it travels. High-sensitivity content executes on local inference; public-grade requests reach cloud reasoning.

  • Automatic redaction

    Deterministic redaction strips identifiers before any cloud call — protection that does not depend on the user noticing.

  • Offline-first, on-device

    Ember Edge — our on-device runtime for Apple Silicon — powers local inference with an offline-first cache. The cloud is a capability extension, not a default.

  • One gateway, full audit

    Every model call crosses one private gateway with provider keys, rate limits and an audit trail in a single place. App code never touches a vendor SDK.

Classify. Redact. Route. Answer. Log.

  1. Prompt arrives

    From the native iOS / macOS client — online or offline.

  2. Sensitivity classification

    Content is classified automatically; the tier decides everything downstream.

  3. Redaction gate

    Identifiers are stripped deterministically before any cloud egress.

  4. Routed execution

    Sensitive tiers run on local Ember Edge inference; public-grade requests reach Ember Reason or Ember Core through the gateway.

  5. Audited response

    The reply returns with the full invocation recorded — model, route, and tool calls via MCP.

Case study

Anonymized professional services teamAnonymized pilot
Challenge
Mixed-sensitivity workloads (some traceable to cloud, some restricted) and no single view over local and cloud model usage.
Solution
CoreAI Hub routed routine prompts to local Ember Edge models on Mac Studio M3 hardware and selectively escalated complex reasoning to Ember Reason with full audit logs.
Outcome
Costs cut against equivalent cloud-only setup; sensitive data stayed on-device; team adoption stuck after two weeks.
Cloud cost vs baseline
−45%
Local-handled prompts
~70%

The private AI stack underneath.

  • Private LLM gateway

    A single internal gateway brokers every model call across Coreidea platforms — keys, rate limits, audit in one place.

  • Local inference engines

    Ember Edge on Apple Silicon delivers low-latency, on-device inference; data stays local by default.

  • One owned gateway, many routed models

    Every request goes through Coreidea’s own gateway: identity, routing, budget and trust labelling are ours. The gateway then routes to the model best suited to the task, including third-party providers, and each response says which model answered.

  • MCP tool integration

    Structured tool calls via the Model Context Protocol make every AI action attributable and auditable.

Where CoreAI fits.

  • Operating teams

    Day-to-day AI assistance over contracts, finances and internal documents — without those documents leaving controlled infrastructure.

  • Invited partners

    Invite-only access for partner organizations that need bilingual private AI with audit guarantees.

  • Research collaborators

    A controlled environment for evaluating sensitivity-aware routing and local-first inference patterns.

  • Apple-native workflows

    First-party SwiftUI clients for iOS and macOS, designed for people who live on Apple hardware.

Questions, answered.

Why is CoreAI invite-only?

CoreAI is live and access-controlled. We onboard operating teams and partners deliberately so that routing policies, redaction rules and audit workflows are tuned to each team’s real usage.

What stays local and what goes to the cloud?

Every prompt is classified first. High-sensitivity content executes on local Ember Edge inference. Only public-grade requests — after deterministic redaction — reach Ember Reason or Ember Core through the private gateway.

Which platforms are supported?

Native SwiftUI clients for iOS and macOS, backed by the private gateway. The clients are offline-first: local inference and cache keep working without connectivity.

How do we get an invite?

Email info@coreidea.cn describing your team and use case. Partner and research-collaboration requests are prioritized.

Give your team AI it is allowed to use.

Request access — sensitivity-aware routing, automatic redaction and full audit, on Apple-native clients.

All platforms