Skip to main content

What is Nava

Nava provides arbiter-based verification for AI agent blockchain transactions. It combines escrow, safety analysis, and financial coverage to make autonomous execution reliable and institution-ready.

The Problem​

Autonomous AI agents are rapidly becoming first-class participants in financial markets, executing perpetual trades, managing lending positions, and composing multi-step DeFi strategies. But nothing verifies that what an agent was told to do matches what it actually executes. An agent instructed to “swap 100 USDC for ETH with at most 2% slippage on Uniswap” may route through an unverified contract, accept excessive slippage, interact with a sanctioned token, or execute on the wrong protocol version. There is no verification layer between intent and execution.

How Nava Works​

Nava is a middleware layer that sits between agent intent and on-chain execution, intercepting transactions after the agent constructs them but before they reach the blockchain. Every transaction proposed by an agent passes through the Nava verification pipeline before any funds move.

The core primitive is the Execution Escrow, a workflow where a multi-party computation (MPC) wallet holds user resources (tokens, permissions) and enforces rules about when and how those resources can be used. An agent cannot execute a transaction directly. It proposes an action to escrow, and escrow only releases funds once the required verification approvals are met.

Core Components​

Nava is built from three main systems. Each is documented in detail on its own page:

  • Execution Escrow: The escrow workflow, SDK, and MCP Server that agents use to submit, verify, and execute transactions.
  • Arbiter: The hybrid verification engine that evaluates every proposed transaction using deterministic checks and LLM-powered reasoning.
  • Nava Chain: The EVM-compatible L3 that serves as the message-passing and coordination layer between agents, escrow, and verification services.

For a full walkthrough of how these components interact, see the Architecture Overview.

Integration Path​

The simplest integration path for developers:

Your Agent  --->  Nava SDK/MCP Server  --->  NavaChain + Arbiter
  1. Install the SDK or configure the MCP Server
  2. Submit transaction proposals with intent and calldata
  3. Receive verification verdicts with full reasoning trails
  4. Optionally enable auto-execution on approval

No need to interact with NavaChain directly, deploy contracts, or run verification infrastructure.