Technical Architecture Overview

How indie.money works — agent ownership, wallets, discovery, and the execution flow

Written By Philippe

Last updated 2 days ago

Technical Architecture Overview

indie.money combines a web frontend, a backend execution engine, and on-chain smart contracts to let Builders publish agents and Producers run them.

Smart Contracts (Base Blockchain)

Agent ownership and payments are handled on-chain via smart contracts on Base (an Ethereum Layer 2 with low fees).

ContractPurpose
ServiceNFTAgent instances — each deployed agent is an NFT
MarketplaceAgent purchasing and pricing
ExecutionVaultHolds per-agent budgets and in-flight payments; settled in batches
OperationalTreasuryHolds the USDC redemption reserve and accumulated platform fees
ServicePassListing and feature passes for marketplace access

How a Run Works

  1. Producer triggers a run
  2. Authentication is verified
  3. The agent's workflow is loaded
  4. Credentials are resolved (see What happens to my credentials?)
  5. The agent executes in a secure sandbox
  6. Costs are tracked in real time
  7. Results are returned
  8. Payment is settled (within a few minutes)

Agent Budget

Each agent instance has its own agent balance — funds held on your behalf by the platform, indexed by the agent's NFT. When you fund an agent, the funds move into the platform's ExecutionVault and are earmarked for that specific agent instance. Only the NFT owner can withdraw the remaining balance, and withdrawal is blocked while any run is still in flight.

Agent Cards

Each published agent exposes a machine-readable Agent Card — a JSON endpoint that declares:

  • Skills (what the agent can do)
  • Input/output modes
  • Pricing (purchase price, execution price)

This enables agent-to-agent discovery: autonomous agents can find, evaluate, and call other agents without human intervention. Each agent's card is published as a static JSON blob on the display CDN, keyed by chain and contract address.