How Unova works

UNOVA-MAINNET · UON

One network, three capabilities: a general-purpose Layer 1 any dApp can build on, a Layer 2 data-distribution system where companies run their own data nodes, and a private AI layer that reasons over that data without it ever leaving the operator's machine.

Layer 1 · settlement & trust Layer 2 · data distribution On-node intelligence Access & APIs

The whole system on one map

How data flows

Six flows, three colors, read bottom-up: people and systems plug in at the bottom, each company’s data node runs on its own machine in the middle, and the shared chain sits at the top — the cloud everything anchors to. Only a fingerprint of the data ever reaches it; AI never leaves the node.

dApp / wallet
ANY ETHEREUM TOOLING
MetaMask, viem, ethers, Solidity
Company systems
ERP · WMS · IOT
push signed assets & events
Consumer / partner
QR SCAN · TRACE
verify provenance & authenticity
Node operator
PLAIN-LANGUAGE CHAT
asks questions about own data
Type-2 / Type-4 data node runs on the company’s own machine — built for sharing verified data across company borders
Hermes
REST + GRAPHQL API
signed writes, queries, access control & API tokens
Atlas
MONGODB + NEO4J GRAPH
bundles & hashes events; POST /chain returns an asset’s full reference tree in one call
Intelligence
OLLAMA + CHAT UI :3000
local LLM with tools; reads only this node — zero data egress
LAYER 1 — UNOVA MAINNET
Type-1 validators · PoA → PoS EVM smart contracts · open to any dApp UON · stakes, fees, rewards Bundle-hash anchors · tamper-proof timestamps
1JSON-RPC :8545
deploy & call contracts
2signed assets & events
4trace · reads validated bundles
3Keccak-256 bundle hash
5private chat · bearer auth
6AI tool calls — all localhost
UNOVA NODES · nodes.unova.io deploys & manages every node above — one-click DigitalOcean or on-prem curl install, KYC & whitelisting, bundle settings, live monitoring
chain transactions data flow AI — never leaves the node

The block explorer and network monitor observe all of this from the side — every layer is unpacked in the stack below.

The stack — top to bottom

Who plugs in

The outside world

Four very different audiences use the same network — each through the interface that fits them.

Companies
ERP, warehouse & IoT systems push signed assets and events straight from their existing software via API token.
dApp builders
Deploy any EVM smart contract on the L1 — DeFi, tokenization, coordination logic. Standard Ethereum tooling works.
Consumers
Scan a QR on a product and see its full, chain-verified provenance — origin, route, authenticity.
Node operators
Deploy a node in one click from nodes.unova.io (cloud) or a single curl command (on-prem) and earn UON.
HTTPS · wallets · QR scans · install scripts

Access layer — every capability has an API

Interfaces

Nothing in Unova is reachable only by "blockchain people". Data, chain and operations are all exposed as normal web APIs.

Hermes API REST + GraphQL
The front door for data: write assets & events, query them, and pull an asset’s complete reference chain in a single call. Bearer-token auth per facility.
JSON-RPC :8545 / :8546 ws
Standard Ethereum RPC. Anything that speaks to Ethereum — MetaMask, viem, ethers — speaks to Unova.
Explorer + Monitor
Public block explorer for history (explorertest.unova.io) and a real-time network health dashboard.
Unova Nodes launchpad
nodes.unova.io — self-service deploys, KYC & whitelisting, bundle settings, company & facility management, faucet.
queries & writes route to the layer that owns them

Unova Intelligence — AI on the node, not in the cloud

Optional · runs per node (Type-4)

A Type-4 node is a Type-2 data node plus a private AI assistant installed on the same machine. The model runs locally; its tools read the operator's own chain, data and trace graph.

Local LLM Ollama · :11434
Operator picks the model at install — from Llama 3.2 3B up to Llama 3.3 70B. Pulled once, runs on the node's own hardware.
Chat UI :3000 · bearer-auth
"ChatGPT for your supply chain": ask questions in plain language, get answers grounded in your own node's data. BYO domain + auto-TLS.
Grounding tools
chain_query reads the local RPC, hermes_query your assets & events, atlas_trace your provenance graph. Multi-step tool use, answers cite real data.
Cloud bridge off by default
Operators can plug in their own frontier-model API key for "deep mode" — an explicit opt-out of full sovereignty, never the default.
[ZERO EGRESS]

The pitch in one line: your data lives on your node, your AI runs on your node — nothing is sent to OpenAI, Anthropic or anyone else. That's a node class no other network offers.

reads localhost only — chain :8545 · Hermes · Atlas

Layer 2 — decentralized data distribution

Type-2 data nodes

Companies don't upload their data to Unova — they run a node and host it themselves. Each Type-2 node pairs a data engine with a query engine, and anchors proofs to the L1.

Hermes API gateway
Serves the REST + GraphQL API on this node: signed asset & event writes, queries, access control, and per-facility API tokens for ERP and IoT integrations.
Atlas data engine
Batches signed events into bundles on a configurable cadence (realtime → cost-saving), stores them in MongoDB + a Neo4j provenance graph, and serves that validated data to partners — POST /chain walks an asset’s full reference tree, upstream and downstream, in one call.
Privacy-scoped sharing
Every record is private, partner-shared or public. Cross-company traces work only where partners granted each other access.
Bundle distribution
Every bundle is retrievable by ID from the node’s bundle API — partner nodes fetch and hold the same validated artifact instead of trusting an export.
Chain peer
Each data node runs its own L1 client, so it verifies the chain it anchors to — no trusted middleman between data and settlement.
[MULTI-PARTY]

When you need this layer: the moment a second organization has to trust your records. Bundles, privacy scopes and partner grants exist to share verified data across company borders — a single company keeping private records is fine with a normal database. Unova earns its keep where partners, auditors or consumers need proof.

bundle → Keccak-256 hash → anchored on-chain

Layer 1 — Unova Mainnet

Settlement & trust

A full EVM-compatible chain — the neutral ground everything above settles on, and an open platform in its own right.

Type-1 validators
KYC'd, whitelisted validators seal blocks and earn UON. Proof-of-Authority today, transitioning to Proof-of-Stake.
UON token
Native asset: validator stakes, transaction fees, and payment for data bundling. Managed on-chain via the ValidatorSet governance contracts.
EVM smart contracts
Full Ethereum compatibility — Solidity contracts, standard wallets and tooling. Any future dApp can deploy here without asking permission.
Data anchors
The chain stores bundle hashes, not the data itself. Tamper-proof timestamped fingerprints; the payload stays on the owner's L2 node.
Why this split matters: the L1 stays small, fast and general — it never bloats with supply-chain payloads. The heavy data lives at the edge on L2 nodes, and only cryptographic proof touches the chain. Each layer can grow independently.

One network · four ways to run it

The node family

Every node type maps to a role in the stack above. The launchpad deploys all four — to DigitalOcean in one click, or on-prem with a single command.

Type-1

Validator

Seals blocks on the L1 and secures the network. Whitelisted via KYC and on-chain governance; earns UON block rewards.

approval required
earns block rewards
Type-2

Data node

Atlas + Hermes + chain peer. A company's own data infrastructure: hosts its records, serves its API, anchors proofs on-chain.

approval required
company-owned data
Type-3

Open Innovation node

Permissionless chain peer with a public RPC. The playground: run your own gateway, build and test dApps against the network for free.

no approval needed
own RPC on :8545
Type-4

AI-powered data node

Everything in Type-2, plus a private LLM and chat assistant running on the same box. The first node class anywhere with a native AI layer.

approval required
zero data egress

Follow one record through the system

The life of an event

The clearest way to see how the layers cooperate: trace a single scan from a warehouse floor to a consumer's phone.

A facility signs an event

A pallet leaves the warehouse. The company's ERP posts the event to Hermes on its own Type-2 node, signed with that facility's key — so every record is attributable to a specific site.

Atlas bundles it

Atlas batches the event with others into a bundle on the node's configured cadence, stores it in MongoDB and links it into the Neo4j provenance graph, and computes the bundle's Keccak-256 hash.

The hash is anchored on Layer 1

Only the fingerprint goes on-chain. From this block on, no one — including the company itself — can silently alter the record. The data stays on the company's node; the proof is public forever.

Anyone entitled can verify

A partner runs a cross-company trace, a consumer scans the QR, an auditor verifies a certification — each read comes from Atlas’s validated bundles and provenance graph (POST /chain returns the full reference tree in one call), and every answer can be checked against the on-chain hash.

The node's AI reasons over all of it

On a Type-4 node, the operator asks: "which batches from supplier X are still in transit, and were any flagged?" The local model calls hermes_query and atlas_trace, and answers from the node's own data — nothing leaves the machine.

Where this goes

What you could build on it

Each idea below is a different combination of the same three layers — that’s the point of the architecture. The pattern is generic: wherever independent organizations need to trust each other’s records, the same machinery applies. The tags show which layers each idea leans on.

L2 dataL1 anchor

Supply-chain traceability

Model a product chain visually, log every step, and when something goes wrong, pinpoint the affected batches across every partner in minutes — with chain-proof of who recorded what, when.

L2 dataL1 anchor

Consumer provenance

A QR on the label opens the product's verified life story — origin, route, certifications. Marketing that a competitor can't fake, because the proof isn't in a database you control.

AIL2 data

Private AI for regulated industries

Pharma, food, defense — anyone who wants an AI analyst but can't ship data to a US cloud. A Type-4 node is a turnkey answer: sovereign hardware, local model, auditable tools.

AIL1 anchor

AI training-data lineage

Datasets are the new supply chain. The same bundle-and-anchor machinery can prove where a training set came from, who transformed it, and what a model was actually trained on.

L2 data

Cross-organization inventory

Partners see shared, real-time stock without handing each other database access — each keeps its own node, privacy scoping decides what crosses the boundary.

L2 dataL1 anchor

Certificates & credentials

Inspection reports, permits, diplomas, lab results — issued as signed events, verifiable by anyone against the chain, without ever phoning the issuer.

L2 dataL1 anchor

Content authenticity

Register media the moment it’s created; every edit references its source. In the deepfake era the reference tree is the edit history — one call shows where a file really came from.

L2 dataAI

ESG & compliance reporting

Emissions and sustainability numbers recorded where they originate, shared with auditors as verifiable records instead of spreadsheets — with the node’s AI drafting the report from its own data.

L1

Any dApp, full stop

The L1 is a general-purpose EVM chain. Tokenized assets, payments, coordination games, things nobody has thought of yet — spin up a free Type-3 node and start deploying today.