Aave Labs said it has launched an official Model Context Protocol server that allows AI agents to query Aave protocol data and prepare transactions across Aave V3 and V4.

The new server is available at mcp.aave.com and is designed to connect AI applications to Aave’s lending infrastructure. According to Aave Labs, the MCP server supports both protocol data queries and transaction preparation functions tied to Aave V3 and V4.

Aave said users can query information including chains, markets, reserves, interest rates, supply and borrowing volumes, and protocol risk parameters. The server also supports wallet-level checks such as account health, supply and borrow positions, claimable rewards, and transaction history.

On the transaction side, the tool supports actions including supply, borrow, withdraw, repay, collateral adjustment, emode, reward claiming, and liquidation. Aave Labs said the server does not directly control user assets. Instead, it returns unsigned transaction requests to a transaction preparation tool, and the user’s wallet signs the final transaction.

The launch also includes a pre-transaction simulation feature intended to let users review expected account health before preparing a transaction. Aave described the release as part of a broader effort to expand how the protocol can be accessed through AI tools, with additional support and functionality planned.

Why It Matters

The release adds a new interface layer between DeFi protocols and AI software. Instead of limiting protocol access to wallets, dashboards and direct developer integrations, Aave is packaging data access and transaction preparation in a format that AI tools can use more easily. That could make protocol interaction more conversational and more programmable without removing user-side signing requirements.

It also points to a broader competition around AI-native crypto infrastructure. For lending protocols and other on-chain applications, the question is no longer only whether they have liquidity or users, but whether their systems can be accessed cleanly by the AI products that developers and traders already use.