The problem Nebula solves

AI tools in DeFi are mostly focused on trading, swaps, and price analysis. Lending and borrowing remain largely unsupported. There is no structured intelligence layer that allows AI agents or users to easily analyze lending markets across protocols.

Lending data is fragmented. Supply APY, borrow APR, liquidity, utilization, and vault strategies are spread across multiple protocols and interfaces. Users must manually check Aave, Morpho, Compound, Euler, and other platforms to understand where capital should be allocated.

This creates two problems.

1/ First, users cannot quickly identify the best lending or borrowing opportunities.

2/ Second, AI agents cannot reason about lending markets because the data is not normalized or easily accessible.

Nebula addresses this by creating an AI-first data aggregation layer for DeFi lending and borrowing.

Nebula aggregates live market data across major lending protocols and presents it in a structured interface that both humans and AI agents can query.

Nebula does not attempt to automate all capital allocation. Instead it provides the intelligence layer needed for informed decisions.

Users and agents can:

• Query the best lending markets for an asset
• Compare supply APY and borrow rates across protocols
• Analyze vault strategies and liquidity conditions
• Inspect onchain positions for any wallet or ENS
• Research lending opportunities across chains

The interface also allows execution through the 1delta aggregator, but the core focus of Nebula is market intelligence and research.

This makes Nebula useful as:

• A DeFi research tool
• A lending market discovery engine
• A structured data layer for AI agents operating in DeFi

Challenges I ran into

The main challenge was normalizing lending market data across different protocols.

Each protocol exposes data differently. Aave and Compound provide lending pools with supply and borrow rates. Morpho introduces vaults and peer-to-peer matching. Some strategies follow ERC4626 vault standards while others rely on custom contract structures.

To make the data usable by AI systems, we built a unified schema for lending markets, vaults, and borrowing conditions. This required aggregating data from multiple APIs including the 1delta aggregator, Morpho Blue GraphQL endpoints, and market data sources.

Another challenge was building a query interface that works well for both humans and AI agents. Natural language commands such as "best USDC yield" or "ETH lending rates" must map to structured queries against the aggregated market dataset.

We solved this by building a command parsing layer that converts natural language prompts into market queries.

Execution support was implemented through the 1delta aggregator. This allows deposits, borrows, and leveraged strategies to be simulated and executed through a single atomic transaction.

Safety was a priority, so all execution flows include a simulation step before transactions are sent to the wallet.