The intelligence layer of Meltiverse · In development

One brain,every product.

MeltBrain is the shared intelligence behind Meltiverse products: research, writing, vision and design judgment, delivered as proposals that each product checks and approves before anything happens.

ServesStudio · Meltify · Customelt · PAKPROMArchitecturePrivate gateway
01Request
02Context
03Expert
04Proposal
INTELLIGENCE
LAYER
Product asksExpert reasonsProduct decides

01 · Definition

Not another AI feature.
One intelligence layer.

+ What it is

A shared runtime every Meltiverse product calls through one contract.

Products call MeltBrain when work needs judgment rather than fixed rules. MeltBrain picks the right expert, works inside the limits the product sets, and returns a result the product can verify. The product stays in charge.

− What it is not

  • A chatbot or a customer-facing app
  • A feature locked inside one product
  • A generic agent platform
  • A replacement for human approval
  • A foundation model of its own

02 · System relationship

Shared intelligence,
local authority.

Each product keeps its data, workflows, approvals and publishing. MeltBrain holds the model providers, the experts and the cost controls, and answers through one private gateway, so every product gets the same safeguards.

01ProductRequest + limits
02MeltBrain
Gateway
Identity · budget
03ExpertDomain task
04ProductValidate & approve
05OutcomePublished, made, sent
System boundaryOnly product backends call MeltBrain, never browsers or storefronts. MeltBrain never writes a product database, and provider keys exist only inside it. A proposal changes nothing until the product validates it and a person or the product's own policy approves it.

03 · Expert domains

Four domains.
One runtime.

Experts are versioned tasks with typed inputs and outputs. They share the same providers, safeguards and cost controls, and each one is added only when a product has a real use for it.

01

Creative · Studio

Research, scripts and shot plans.

Episode research with cited sources, story drafts and production plans that Studio's gates approve or reject.

Next: research expert
02

Design · Meltify

Bounded design proposals.

Typography and placement proposals inside a pinned Design Contract. Meltify revalidates every operation and a person reviews it.

Contract v1.0.0 ready
03

Production · Customelt

From a LINE chat to a job brief.

Draft briefs from customer conversations and read payment slips for staff to confirm. Shops keep the final say.

Moving onto MeltBrain
04

Property · PAKPROM

Operations that explain themselves.

Help for dorm and apartment owners where judgment saves time, with every decision staying in PAKPROM.

Planned

04 · Built in

Safeguards
by default.

Models and providers will change. What stays is how every call is handled: who may ask, whose data it touches, what it may cost, and what happens when something goes wrong.

01

Verified callers

Each product calls with its own service identity. A product is never taken from a request body.

02

Tenant isolation

Every call is scoped to a product and its tenant. Nobody can read another tenant's work.

03

One request, one charge

Resending the same operation returns the same result instead of paying for it twice.

04

Budgets before spend

Cost is reserved before any provider is called and settled from actual usage.

05

No silent retries

When an outcome cannot be proven, it is marked unknown for a person to reconcile, never resent.

06

Minimal data

No prompts or customer inputs are kept. Short-lived encrypted payloads are deleted when the work ends.

Every callThe same path for every product and every expert.
Accept→Reserve→Run→Receipt→Settle

In development

Building the foundation
before adding experts.

The contract and SDK are in place. Next comes the gateway foundation, then Studio's research expert, Customelt's existing AI moving across, and Meltify's design proposals once its own product gate is met.

Contract v1 · doneGateway · nextFirst experts