# Stratifi.xyz

## Overview

Stratifi.xyz is an AI agent that interfaces with multiple lending protocols to optimize the liquidity and positions of both lenders and borrowers. Lenders can choose the most optimal pool to stake based on their risk and APY preferences. By abstracting away the complexity of protocol interactions, it enables:

- **Lenders** to find the most optimal pools based on their preferences for risk and APY.
- **Borrowers** to unlock maximum liquidity with minimal effort by identifying optimal collateral and borrowing strategies

Stratifi.xyz offers users the flexibility to choose their preferred level of control. They can opt for full autonomy, where the agent handles all transactions automatically and only displays the final investment outcomes, or they can use it as a research and decision-support agent, maintaining more hands-on involvement. This flexibility significantly lowers the entry barrier for users in decentralized finance (DeFi), making it more accessible and intuitive while ensuring efficient decision-making.

Stratifi.xyz addresses DeFi's accessibility challenges and fragmented user experience, making **intelligent**, **autonomous** **finance** the new standard.

## Purpose and Rationale

In the current DeFi ecosystem, users face significant challenges:

**Complexity of Choice:**
- With hundreds of protocols and pools, finding the most suitable one based on individual preferences (risk tolerance, desired APY, etc.) is overwhelming.
- Borrowers often struggle to determine which assets unlock the highest liquidity for the least collateral.

**Dynamic and Volatile Markets:**
- DeFi is inherently fast-paced and data-heavy.
- Manual decision-making is not only time-consuming but also prone to inefficiencies in such a dynamic environment.

**Fragmented Ecosystem:**
- Users must often switch between platforms, wallets, and analytics tools to execute even simple strategies.
- Stratifi.xyz consolidates this into a unified interface powered by AI.

**Accessibility Barriers:**
- Novice users often find the steep learning curve intimidating, which deters adoption.
- Stratifi.xyz simplifies the process, making Leverage Defi accessible to a wider audience without requiring advanced knowledge of blockchain.

## Composable Autonomy of Agent

We give an option for the user to configure the level of human interaction needed for the agent to fully execute desired strategy, be it borrowing or lending.

**Fully Autonomous** - One step interaction from user to approve fund utilization

**Partially autonomous** - User will have to confirm each transaction decided by the agent. The decisions are presented in a human readable format so the user can decide the next step

**Zero autonomy** - This will act as a research tool for DeFi users. They can use suggestions provided by the model to make decisions but the agent itself will not interact with the chain on behalf of user

## Design

Stratifi.xyz consists of four important components:

### Scout Engine
The Scout Engine is a user-facing component that takes inputs of preferences and efficiently checks on-chain data from various protocols to provide the most optimal position. It considers parameters like risk preference, TVL, APY, and collateral threshold for lending positions. For borrowers, it takes inputs such as available collateral, borrow time, and volatility threshold. This way, users don't have to navigate individual platforms for the best position; everything can be done from one place.

### Execution Smart Contracts
These contracts interface with different lending protocols like AAVE, Compound, and Lido. They also act as vaults for collateral and returns, which users can withdraw. Additionally, they allow composable lending and borrowing within the Stratifi.xyz platform and interface with some off-chain metrics based on user choices. The agent communicates with this smart contract and this contract is responsible for payments of gas.

### Data Aggregator
The Data Aggregator indexes and processes large amounts of on-chain data, both real-time and historical, to derive the spectrum of options for a user's choice. It employs various indexing protocols combined with optimization and filtering techniques.
Major data sources like Messari subgraphs, on-chain parameters are collected and processed directly from the blockchain.

### Think Tank - Composable ML Models
#### Core Functionalities
**Data Aggregation and Processing**
- Extensive Data Sources: The Think Tank aggregates data from multiple channels, including on-chain metrics, subgraphs, and market indicators.
- Preprocessing Pipelines: Data is normalized, cleaned, and enriched to address missing values, scale variations, and noise, ensuring optimal input quality for model training and inference.

**Model Selection for Use Cases**
- Stratifi.xyz employs tailored ML models for specific financial scenarios:
- Borrowers: Receive forecasts on future interest rates, liquidity levels, and optimal borrowing strategies.
- Lenders: Gain access to risk classifications, yield predictions, and recommendations for optimal staking pools.
- Dynamic Composability: Models are modular, allowing seamless integration, updates, or replacement based on evolving requirements or new insights.

**Parallelized Model Execution**
- Multi-Model Strategy: The Think Tank runs multiple models in parallel, each targeting distinct objectives, such as risk assessment, interest rate prediction, or liquidity analysis.
- Holistic Insights: This parallel approach mitigates individual model biases and delivers a comprehensive market analysis.

**Optimization and Continuous Learning**
- Stratifi.xyz's ML models undergo continuous training and fine-tuning, incorporating new data streams to adapt to the rapidly evolving DeFi landscape.
- Real-Time Adaptability: The Think Tank evolves its recommendations in sync with market dynamics, ensuring relevance and accuracy.

**Actionable Outputs**: These results are human readable and compatible for onchain execution.

| **Model Type** | **Objective** |
|----------------------------|---------------------------------------------------------------------------------------------------------------------------------------------------|
| **Time Series Prediction** | To forecast temporal sequences of data points, such as protocol usage metrics, asset prices, or wallet transaction volumes. For example AAVE. |
| **Classification** | To categorize data points into discrete classes based on learned features. Essential for tasks like risk assessment, user segmentation, and behavior prediction. |
| **Regression** | To predict continuous outcomes based on a set of input features, helping to understand relationships between variables and make quantitative predictions. |

## User Flow

### Asset Management Protocols and Individual Lenders

We provide a simple plug and play Widget or extension where users can input the asset holdings and how they want to invest in a human understandable manner.

For example: "I have 100 USDC and would like to know what is the highest APY I can get and should have higher liquidity with low risks"

Or with a wallet connected:

"Invest 50% of cold funds in high yield, high risk pools."

Or they can simply use the extension to compare the returns they are getting in a Dapp to other pools in the same Dapp and swap assets for higher returns or security.

**Automate** - Automate the rebasing of assets by providing allowance for specific assets up to specific spending limits and allow the agent to manage portfolios.

Asset management protocols can utilize a combination of these ML models to modify how the AI agent behaves. This way they can fine tune their AI agent without having to deal with the complexity of training complex models.

**Custody-free delegation of assets**

One need not deposit tokens to these AI agents. Instead, we allow delegation approach where the investor can define the bounds of investment and make sure the agent behaves within those parameters.

**Stable Vaults**
-Stable Vaults: Modified ERC4626-like vaults that deposit user assets (e.g., USDC, STRK, ETH) into money markets (zkLend, Nostra, Vesu) based on the highest supply-side yield, with a 2% fee.


**Dynamic Vaults**
- CLMMs allow positions to be concentrated across multiple fees and tick tiers. Our dynamic vaults distribute liquidity across these tiers for a single pair. For example, for the ETH-USDC pair, Stratifi.xyz opens the major position within the (0.05%, 0.1%) range and holds smaller positions in the other two ranges. This approach ensures that if there's a delay in rebalancing, the minor positions continue to accrue fees, minimizing losses and maximizing yield.

**Options Vaults**
The vault buys equal quantities of put and call options at the same strike price and expiration date for highly volatile assets, known as the straddle strategy. This benefits the vault and its high-risk, high-reward users, as the vault can exercise the more profitable option regardless of which way the price moves. However, the risk involved is that if the asset's price doesn't move significantly, both options could expire worthless, resulting in a premium loss. Additionally, the strip strategy, which involves buying one call option and two put options, increases downside protection, providing a larger profit if the price drops while maintaining limited risk if the price rises. The vault will utilize Carmine options and an off-chain bot to continuously monitor and signal the vault for potential yield opportunities, automating the process.

**Proactive Market Maker**
The proactive market maker operates through a combined on-chain and off-chain mechanism. Liquidity providers deposit any amount directly into the respective vault, while the off-chain component monitors deposits and withdrawals to manage its inventory. When a swap request is made via our user interface or aggregators like AVNU or Fibrous, the API queries the off-chain service for quotes.

The off-chain service returns a signed signature with the quote, which is then verified and settled on the on-chain router. The router encapsulates the specific vault where liquidity was deposited, validates the signature, and executes the trade at the quoted price with zero slippage. This system benefits all parties involved: users enjoy zero-slippage trades, liquidity providers earn competitive fees, and impermanent loss is minimized. Similarly, off-chain market-making algorithms can be fine-tuned to adapt to market conditions, introducing dynamic fees during periods of high volatility and unlocking numerous additional possibilities.

### Unlocking Optimal Liquidity for Borrowing Parties

The tool plays a huge role in unlocking liquidity for borrowers. Ideally one would choose to borrow from pools which provide maximum liquidity with lowest collateral and lower interest rates.

The agent will return the best pools to borrow from based on the asset availability in the user's wallet. Stratifi.xyz evaluates collateral efficiency (borrowable amount vs. asset value) and borrowing costs (interest, fees, etc.). If not ideal, the agent will give the user optimal pools and also offer to execute the transaction on behalf of the user.

## Programmable Agent Hooks for Advanced DeFi Workflows

Developers can program our agents to interact with other AI agents, passing parameters for any follow-up actions triggered by liquidity provisioning or unlocking. Developers can program our agents to call other AI agents with any parameters to do any follow up actions after providing or unlocking liquidity.

These hooks are categorized into two types:

**PreX Hooks** (Pre-Execution Logic):
These hooks define actions or checks that occur before the primary transaction is executed. For instance:
- Validating user-defined conditions before borrowing assets.
- Fetching additional market data to refine execution parameters.

**PosX Hooks** (Post-Execution Logic):
These hooks facilitate actions that occur after the main transaction. One can call other agents for subsequent operations.

### Example Use Cases for Hooks
**Automated Swaps Post-Borrowing:**
A user borrows assets and configures the agent to delegate the tokens to a swapping agent. The swapping agent identifies the most profitable pool and executes the swap to the desired asset.

**Staked Token Utility Enhancements:**
Users can configure post-staking actions, such as delegating staked tokens to other protocols for additional yield opportunities or enabling cross-protocol operations (e.g., staking on one chain and using a bridged version of the staked token on another chain).

## Token and its Utility

- **Platform fee** - Users can use STRAT token to buy BIPS which allow users to use the agents. BIPS can be earned by staking STRAT tokens or by purchasing them using STRAT.
Staking refreshes BIPS quota of user on a daily basis.
- **Single gas token for all chains** - The AI agent uses gas fee to interact with all the supported chains and lending protocols. Making sure the agent is always funded to execute transactions in each chain is not possible. Hence we allow you to use our token as the gas fee token to compensate these agents.
- **Additional lending rewards** - Staking certain supported tokens via our Agent will result in rewards in our token along with the APR generated in the lending pools.
- **Subsidized interest rates** - High volume long term holders of our tokens will get subsidized interest rates for borrowed assets.
- **Community participation/Governance** - The agent kit will initially support only a few chains and protocols. The community can choose to add new chains and train ML models on new protocols and choose to publish to the main kit. This will be vetted by the governance participants and based on consensus will be added to the platform.

## Roadmap
# Stratifi.xyz Roadmap

| **Timeline** | **Milestone** | **Key Deliverables** |
|--------------|---------------|----------------------|
| **Q1 2025** | Foundation Building | • Complete technical architecture and specifications<br>• Begin Scout Engine prototype development<br>• Establish initial protocol partnerships (AAVE, Compound, Lido)<br>• Start data collection pipeline for Ethereum & Polygon<br>• Develop ML model frameworks for lending analysis |
| **Q2 2025** | Core Tech Development | • Launch Scout Engine alpha with basic recommendations<br>• Deploy first ML models for risk assessment & yield prediction<br>• Build protocol interaction smart contracts<br>• Implement Partially autonomous mode<br>• Implement real-time data aggregation layer<br>• Release UI with natural language capabilities<br>• Complete initial security audit |
| **Q3 2025** | Beta Release | • Support 3-5 major lending protocols<br>• Launch on initial chains (Ethereum, Polygon, Arbitrum)<br>• Implement Fully autonomous mode<br>• Deploy stable vaults with single-pool strategy<br>• Release research tool functionality<br>• Begin limited user testing with DeFi power users |
| **Q4 2025** | Expansion & Token Launch | • Expand to 10+ lending platforms<br>• Add support for more chains (Optimism, Base, zkSync)<br>• Launch STRAT token & BIPS credit system<br>• Deploy dynamic vaults & gasless transactions<br>• Onboard strategic partners & liquidity providers |
| **Q1 2026** | Full Production Launch | • Enable fully autonomous mode with risk controls<br>• Implement options vaults & programmable hooks<br>• Launch governance portal & cross-chain operations<br>• Complete comprehensive security audit<br>• Scale marketing for broad DeFi adoption |
| **Q2-Q3 2026** | Advanced Features & Growth | • Deploy AI portfolio optimization & market maker functionality<br>• Launch advanced ML models for market prediction<br>• Implement agent-to-agent communication<br>• Expand to emerging L2s & app-specific chains<br>• Introduce developer grants & customizable agent behavior |

## Revenue Model

Stratifi.xyz generates revenue through the following streams:

### Transaction level Fee
- We charge a 0.01% premium on the volume of each on-chain transaction
- We handle all gas payments across multiple chains, allowing users to execute transactions without needing to hold native tokens (ETH, MATIC, etc.) for gas
- This model provides users with gasless transactions, simplifying their experience across different blockchains

### Platform tiers

**Research Tool**
- Free tier
- Provides analytics and recommendations without executing transactions
- Users can explore optimal lending and borrowing opportunities at no cost

**Autonomous Agent - Action Level**
- Pay-per-use model (postpaid) using STRAT token.
- Users are charged based on the number and complexity of actions processed by the agent.
- Stake STRAT tokens to earn BIPS credits which are needed to run these agents

## Token Metrics (TBD)

- Max Supply
- Initial Supply
- Ticker
- Allocation
- Development
- Public Sale
- Airdrops and grants
- Investors and advisors
- Buyback and burn
- Team