Written by Helena Markou
A new proof-of-concept demonstrates an AI agent that self-funds via Base, processes confidentially on SecretVM, and retains permanent memory through Auto Drive, setting a new standard for crypto infrastructure.
In the rapidly evolving landscape of decentralized finance and blockchain infrastructure, a critical vulnerability has long persisted: the reliance on centralized maintenance. Most decentralized applications and their associated AI agents require human intervention to pay for servers, maintain databases, and secure private keys. The moment this external support ceases, the system halts. However, a groundbreaking proof-of-concept developed by the Secret Network team, in collaboration with Autonomys Network, has successfully demonstrated a self-sustaining AI agent that manages its own operational costs and maintains persistent memory without human oversight.
This development marks a significant milestone for crypto infrastructure, illustrating the powerful synergy between confidential computing and decentralized storage. The agent, aptly named the Funding Agent, operates entirely within a Secret Virtual Machine (SecretVM) on the Secret Network, a mainnet blockchain renowned for its privacy-preserving smart contracts and end-to-end encryption.
The architecture leverages SecretVM’s Trusted Execution Environment (TEE) to isolate computation at the hardware level. This ensures that all data, including private keys generated for the agent’s wallet on the Base network, remains encrypted both in transit and at rest. Because these keys are generated and sealed inside the VM, they are completely inaccessible to any human observer, eliminating a major vector for security breaches in crypto-native applications.
What truly distinguishes this agent is its economic autonomy. Equipped with its own Base wallet, the Funding Agent receives USDC contributions directly from users who interact with it. When its compute credits begin to run low, the agent programmatically transfers USDC from its wallet to its VM balance. This self-funding mechanism requires no external scheduling or triggering scripts; the agent autonomously manages its operational overhead as long as it retains a positive wallet balance. This creates a fascinating micro-economy where the utility provided by the agent directly funds its continued existence on-chain.
However, confidential compute alone is insufficient for a truly autonomous agent; it also requires persistent state. Without a decentralized memory solution, an agent operating in a TEE would lose its entire history upon rebooting. To solve this, the architecture integrates Auto Drive from the Autonomys Network. Every interaction the agent has is recorded as a JSON file and permanently stored across Autonomys’ Distributed Storage Network.
Every interaction the agent has is recorded as a JSON file and permanently stored across Autonomys’ Distributed Storage Network. This decentralized storage layer ensures that whenever the agent boots up or a new VM instance is spun up, it can seamlessly retrieve its full conversational history and resume operations. The integration of Auto Drive provides a permanent, verifiable on-chain record that is completely decoupled from any single machine or centralized server. For builders requiring privacy, Auto Drive also supports client-side end-to-end encryption before upload, ensuring that data remains verifiable on-chain while only accessible to those holding the decryption keys.
The implications for the broader crypto and DeFi ecosystems are profound. By combining Secret Network’s confidential computation with Autonomys’ permanent storage, developers can now deploy agents that execute complex financial strategies, manage decentralized autonomous organizations (DAOs), or provide automated market-making services with absolute privacy and continuity. These agents can operate without exposing their underlying logic or trading strategies to external observers, all while maintaining an immutable audit trail.
As blockchain infrastructure matures, the shift toward fully autonomous, self-sustaining entities represents the next frontier. The Funding Agent is not merely a technical demonstration; it is a blueprint for the future of decentralized computing, where AI agents exist as sovereign economic actors on-chain, managing their own resources and memories in a trustless environment.
