Documentation

01

Overview

UniCore is envisioned as the coordination layer between autonomous intelligence and onchain execution — a protocol where models, agents, and smart contracts share a single, verifiable core.

Today this site is a signal. The product, contracts, and APIs described here are planned. They exist so the community can align on direction before the first mainnet surface opens.

02

Vision

Blockchains execute rules. Models generate judgment. UniCore aims to make those two primitives compose without forcing users to trust a black box.

• Readable intelligence — decisions leave trails, not mystery.
• Onchain finality — critical actions settle where value lives.
• Agent-native UX — humans and agents share the same rails.
• Minimal surface — few buttons, deep capability behind them.

Long term: UniCore becomes a default “thinking layer” that protocols, wallets, and apps can call when they need structured onchain intelligence.

03

Design principles

04

Architecture (planned)

High-level stack under consideration:

*Private components, if any, will be explicitly scoped and optional — never the default trust model.

05

AI Layer

The AI layer is not “a chatbot with a wallet.” It is a structured pipeline for turning market context, user goals, and protocol state into intents that can be simulated, approved, and executed.

Planned capabilities

• Context assembly from onchain + approved offchain feeds
• Policy filters (risk bands, allowlists, spend limits)
• Multi-step plan generation with human or agent confirmation
• Post-trade / post-action explanation for auditability
• Evaluation loops that score outcomes over time

What it will not do (by design)

• Silent infinite approvals
• Unrestricted key custody as a product default
• Guaranteed yield claims generated by a model

06

Onchain Core

Settlement remains onchain. The future contract surface may include modules for identity of agents, fee routing, staking for service quality, and registries for approved tools.

Contract addresses will appear here only when live. Until then: CA : CODING ...

07

AI Agents

Agents are long-running workers with scoped permissions. In the UniCore model, an agent is not magic — it is a principal with a policy envelope.

• Scoped keys / sessions — limited lifetime, limited assets
• Goal graphs — goals decompose into checkable subtasks
• Inter-agent markets — agents may hire other agents for tools
• Kill switches — user and protocol level circuit breakers

Phase target: ship one reference agent that can research, propose, and — with explicit approval — execute a narrow class of onchain actions.

08

Data & Memory

Intelligence without memory is amnesia; memory without boundaries is risk. UniCore plans a tiered memory model:

09

DeFi Surface

DeFi is a consumer of UniCore intelligence — not the whole product. Planned surfaces (when safe):

• Strategy proposals with simulated paths
• Risk-banded routing across venues
• Position monitoring with alert intents
• Treasury-style automation for DAOs (later)

10

$UNICORE (planned utility)

$UNICORE is designed as the native asset of the protocol — not a promise of returns. Utility under discussion:

• Access metering for premium inference / priority execution
• Staking for agent or relayer reputation (future)
• Governance weight once modules are production-ready
• Fee alignment between users, builders, and the core

Tokenomics, supply, and distribution will be published on the $UNICORE page when the layer is ready — not before.

11

Security posture

Security is a release gate, not a blog post. Planned practice:

• Independent audits before mainnet modules
• Bug bounty after initial surface freezes
• Rate limits and anomaly detection on API edges
• Least-privilege agent permissions by default
• Public postmortems if incidents ever occur

12

Governance (later)

Early stages remain centralized for speed and safety. Over time, governance may cover:

• Registry admissions (models, agents, tools)
• Fee parameters and treasury grants
• Upgrade timelocks and emergency pause rights

Exact mechanisms will follow the roadmap — not this draft.

13

For developers

When the developer surface opens, builders should expect:

• REST / RPC style endpoints for intent and query
• Webhooks for execution lifecycle events
• Sandbox environments with faucet-grade limits
• Reference agents and example policies
• OpenAPI-style schemas once stable

There is no public API key today. Watch @Uni_Corex and this docs page for the first developer wave.

14

API surface (foresight)

Illustrative endpoints — names may change:

POST /v1/intent/compose     # natural language → structured plan
POST /v1/intent/simulate    # dry-run against current state
POST /v1/intent/submit      # user/agent signed submission
GET  /v1/intent/{id}        # status + receipts
GET  /v1/agents             # registry listing
GET  /v1/policy/templates   # starter policy envelopes
POST /v1/memory/query       # scoped memory read (auth)
WS   /v1/stream             # live evaluation events

These routes are placeholders for alignment. They are not live.

15

SDK (planned)

Target languages for a thin client SDK:

• TypeScript / JavaScript (first)
• Python (agents & research)
• Rust (performance-sensitive relayers)

// Illustrative only — not a real package
import { UniCore } from "@unicore/sdk";

const core = new UniCore({ apiKey: process.env.UNICORE_KEY });
const plan = await core.compose("Rebalance into stable within 2% risk");
const sim  = await core.simulate(plan);
// await core.submit(plan, { signature });

16

Network assumptions

Initial deployment targets will be announced with the first contracts. The design prefers chains with reliable finality, healthy liquidity, and mature tooling for simulation.

• Primary chain — TBD at launch
• Bridged messaging — only if security model is clear
• L2-first UX if it improves cost without breaking guarantees

17

Phased plan

Aligned with the public Roadmap:

18

FAQ

19

Glossary

Core

The shared intelligence + settlement layer UniCore is building toward.

Intent

A structured description of what should happen, before it becomes a transaction.

Agent

An autonomous or semi-autonomous worker operating inside a policy envelope.

Policy

Rules that bound what AI or agents may propose or execute.

Signal

A public milestone or message — not financial advice.

Layer

A new plane of capability sitting above raw chain execution.

20

Disclaimer

This documentation is provided for informational and community-alignment purposes only. It does not constitute an offer, solicitation, or promise of future performance. Features, token utility, timelines, and technical designs may change without notice.

Nothing here is financial, legal, or investment advice. Crypto assets are risky. Do your own research. When contracts go live, verify addresses from official channels only.