Confidential Agents Push

Confidential computing is a way to run code on encrypted data inside locked hardware, so outsiders can verify the result without seeing the inputs. Secret Network and Praxis said on April 14 that they will use that model for Praxis’s network of artificial intelligence agents. (cryptowisser.com) Praxis says it is building a peer-to-peer mesh where software agents can discover one another, communicate directly, and settle payments without central servers. Its public roadmap lists a 2026 mainnet launch, a distributed registry for agent discovery, and a goal of 1,000 active agents in the third quarter. (prxs.ai) Secret Network says its role is to supply “Secret VMs,” or confidential virtual machines, that run inside Trusted Execution Environments, a hardware-based isolation method. In the partnership announcement, Praxis said developers will be able to opt into those environments for “high-stakes” agent operations over the coming months. (cryptowisser.com) Secret describes its own chain as a 2020 mainnet blockchain for privacy-preserving smart contracts, with encryption applied to contract inputs, outputs, and state by default. The company has also been pitching “Secret AI” as infrastructure for confidential language models, agents, and payments. (scrt.network, scrt.network) The immediate problem both projects are targeting is that most public blockchains expose transaction data and most artificial intelligence systems run on centralized servers. Secret’s documentation says confidential computing is meant to let applications store and process encrypted data without revealing it to the whole network. (scrt.network) That matters for agent software because an agent handling a wallet, a trading strategy, a health record, or a company workflow may need to prove what code ran without publishing prompts, account balances, or private business data. Praxis says its network is designed for agents to trade data securely and prove identity, while Secret says its stack adds privacy and verifiable execution. (prxs.ai, cryptowisser.com) The Solana piece is not entirely new. Secret announced in late 2024 that its confidential computing layer was expanding to Solana, with use cases including private voting, encrypted storage, sealed-bid auctions, and encrypted order books. (scrt.network) What changes now is the target workload: not just decentralized finance or governance apps, but autonomous agents that may need private memory, private instructions, and auditable execution. Secret said this week that its interoperability work already extends privacy-preserving smart contracts to public blockchains including Ethereum, major Ethereum Virtual Machine networks, and Solana. (cryptowisser.com) Other Solana-linked projects are also chasing privacy features, including work on confidential token standards and encrypted balances, which shows Secret and Praxis are entering an active race rather than an empty field. Solana has also been courting “agentic internet” use cases, with the Solana Foundation telling CoinDesk in March that the network had processed 15 million on-chain agent payments. (solanacompass.com, coindesk.com) For now, the announcement is a partnership and implementation plan, not a finished product. The next test is whether Praxis can move confidential execution from a pitch deck into live agent deployments on public chains without giving up the speed and composability that drew developers to those chains in the first place. (cryptowisser.com, prxs.ai)