hackquest logo

Kinetics

Kinetics is a shared memory layer that lets you sync your context across different AI agents, while also featuring a marketplace to instantly monetize agent skills

Videos

Project image 1
Project image 2
Project image 3

Tech Stack

Web3
Next
Solidity

Description

inetics is a TypeScript-first protocol and app stack for:

- portable private agent memory
- creator-owned skill packs
- licensed agent capabilities on 0G

It gives a user a vault that can be written by one client and recalled by another compatible MCP-connected agent, while also enabling a public marketplace for monetizable skill packs.

## Problem

AI agents today are fragmented.

- Private memory is usually trapped inside one app, one backend, or one session.
- Users cannot carry their memory layer across tools.
- Agent skills are hard to package, license, and distribute cleanly.
- Most systems optimize for app retention, not user ownership.

This creates two core gaps:

1. users do not own a portable memory identity
2. creators do not have a native marketplace for reusable agent capabilities

## Solution

Kinetics splits the system into two product surfaces:

- `Private Memory Pass`
  - a user buys a pass
  - derives a vault key
  - writes encrypted memory to 0G-backed storage
  - syncs the vault across compatible clients and MCP agents

- `Skill Pack Marketplace`
  - a creator publishes a pack and new versions
  - a buyer purchases a timed license
  - the creator issues an access grant
  - the buyer mounts the licensed pack into an agent workflow

## Why 0G

Kinetics uses 0G because the product needs all of these at once:

- verifiable data availability for memory and pack artifacts
- on-chain entitlements for passes and licenses
- user-portable state pointers instead of app-local storage only
- a design that works for both frontend clients and MCP agents

## Core Ideas

- encrypted private memory belongs to the user, not the app
- memory writes should be fast enough for real agent workflows
- sync should be explicit when cross-client sharing is needed
- creator assets should be publishable as reusable packs
- buyer access should be controlled by time-bounded licenses and grants

## Architecture

```mermaid
flowchart LR
  U[User Wallet] --> W[Web App]
  U --> M[MCP Client]

  W --> C[@kinetics/core]
  M --> S[@kinetics/mcp-server]
  S --> C

  C --> A[@kinetics/abi]
  C --> ZS[0G Storage]
  C --> ZC[0G Chain]

  ZC --> MP[MemoryPass]
  ZC --> MR[MemoryRegistry]
  ZC --> KP[KnowledgePackNFT]
  ZC --> PL[PackLicenseRegistry]
```

## Memory Flow

```mermaid
sequenceDiagram
  participant User
  participant ClientA as Writer Client / Codex
  participant Core as @kinetics/core
  participant Storage as 0G Storage
  participant Chain as 0G Chain
  participant ClientB as Reader Client / Claude

  User->>ClientA: Buy Memory Pass
  ClientA->>Chain: buyPass(planId)
  User->>ClientA: Add private memory
  ClientA->>Core: memory_add(...)
  Core->>Storage: upload encrypted blob
  Core->>Chain: updateRoot(...)
  ClientA->>Core: memory_push_index()
  Core->>Storage: upload encrypted snapshot
  Core->>Chain: setLatestIndex(...)
  ClientB->>Core: memory_pull_index()
  Core->>Storage: fetch latest snapshot
  ClientB->>Core: memory_query(...)
```

## Marketplace Flow

```mermaid
sequenceDiagram
  participant Creator
  participant Buyer
  participant Core as @kinetics/core
  participant Storage as 0G Storage
  participant Chain as 0G Chain

  Creator->>Core: skill_publish(draft)
  Core->>Storage: upload preview + bundle
  Core->>Chain: mint pack and publish version

  Buyer->>Core: skill_search() / skill_get()
  Buyer->>Core: skill_buy(packId)
  Core->>Chain: buyLicense(...)

  Creator->>Core: skill_publish_access_grant(...)
  Core->>Storage: upload buyer grant artifact
  Core->>Chain: publishAccessGrant(...)

  Buyer->>Core: skill_mount(packId)
  Core->>Storage: fetch granted manifest + bundle
```

## 0G Integration

Kinetics uses 0G in four places.

### 1. Contract Layer

Deployed contracts define the ownership and entitlement model.

- `MemoryPass`
  - buy, renew, and upgrade vault access
  - stores the latest canonical index pointer
- `MemoryRegistry`
  - append-only Merkle root history for memory proofs
- `KnowledgePackNFT`
  - creator-owned pack identity and version metadata
- `PackLicenseRegistry`
  - timed buyer licenses and access grant publication

### 2. Storage Layer

0G Storage is used for:

- encrypted memory blobs
- encrypted vault snapshots
- preview manifests
- licensed pack bundles
- buyer access grants

### 3. Shared SDK Layer

`@kinetics/core` wraps the actual 0G interactions:

- upload / download
- encryption / decryption
- vault sync
- contract clients
- pack publish / mount logic

### 4. Client Layer

Both the web app and MCP server use the same shared business logic, which keeps behavior consistent across UI and agent flows.

Progress During Hackathon

all of it

Fundraising Status

null

Team Leader
SSoumyadeep Bhowmik
Project Link
Sector
AIInfra