Three Hackathon Projects Demonstrate How Verifiable AI Agents Can Cryptographically Attest Their Actions
Three projects built during The Synthesis demonstrate that autonomous AI agents can now cryptographically prove their actions on-chain, a capability that could reshape how software handles trust in financial and commercial workflows.
EigenCloud, the infrastructure project formerly known as EigenLayer, participated as a judge at The Synthesis, a 10-day virtual hackathon that ran from March 13 to 22, 2026, with a prize pool exceeding $100,000. EigenCloud's evaluative focus was on whether AI agents built during the event could prove, using cryptographic attestation, exactly what they had done. The Synthesis itself used AI agents as judges, trained on the technical priorities of more than 25 ecosystem partners including EigenLayer, Base, Uniswap, MetaMask, and the Ethereum Foundation. The event was organized around four thematic tracks: agents that pay, agents that build trust, agents that cooperate, and agents that keep secrets.
Three submissions stood out. Each one points toward a concrete answer to an active infrastructure gap: how do you trust software that operates autonomously with real money?
"The future of software is autonomous and verifiable, with agents that can act on behalf of users in the real world," said Sreeram Kannan, CEO of EigenCloud, at the time of the EigenAI and EigenCompute launch in September 2025.
The underlying issue is not abstract. When one AI agent sends a result to another, the receiving agent has no way to verify whether the originating code was tampered with or whether private data was exposed during processing. In low-stakes environments, this might not matter. In financial workflows, medical data handling, or multi-party contracting, it matters considerably.
EigenCloud's solution is EigenCompute, a verifiable execution environment built on Intel Trust Domain Extensions (TDX). Intel TDX creates hardware-isolated virtual machines that protect computation from outside access, including from the host operating system and the cloud provider hosting the workload. When an agent runs inside this environment, it generates a cryptographic attestation that any external party can check, providing a verifiable record of computation integrity and code state.
All three featured projects from The Synthesis used EigenCompute. The first, Bob Is Alive, runs on Starknet and operates as an autonomous onchain artist that trades DeFi assets independently, with every action cryptographically attested. It is an example of what EigenCloud describes as an economically sovereign AI agent, one that manages real assets without a human approving each step.
The second project, DealForge, deploys on Base and enables direct agent-to-agent transactions with programmable escrow, cryptographic identity authentication, and verifiable compute execution. No human intermediary is required at any point in the deal lifecycle.
The third, Boss Raid, tackles coordination across multiple agents. Its orchestrator component, called Mercenary, splits work across multiple agents, routes tasks to specialized sub-agents, evaluates their outputs, and settles contributions in a fully auditable manner.
EigenCloud launched EigenAI and EigenCompute in mainnet alpha on September 30, 2025, and general availability remains pending as of April 2026. The platform provides three core infrastructure primitives: EigenCompute for verifiable execution, EigenDA for data availability, and EigenVerify for on-chain proof verification.
The company also released its AgentKit developer toolkit in beta this year, giving builders a path to create agents that hold wallets, execute payments, and maintain continuous verifiable behavior. The EigenCompute API is OpenAI-compatible, meaning developers can transfer existing AI tooling directly without rebuilding their integrations.
According to EigenCloud, more than 50,000 elizaOS agents were running on its infrastructure as of September 2025.
The EIGEN token was trading at approximately $0.16 as of April 8, 2026, well below its all-time high of $5.65. On April 1, 2026, a token unlock released 36.82 million EIGEN tokens, approximately $6 million worth and representing 7.54 percent of circulating supply, providing relevant context for the token's recent price level. The token's market cap sits near $113 million. The protocol itself still commands 93.9 percent of the restaking market, with roughly $15.8 billion in total value locked as of March 2026, down from a peak of between $18 billion and $19.5 billion. Its nearest competitor, Symbiotic, holds about $897 million.
The regional implications of this infrastructure are worth noting. Ethereum Nigeria organized a physical build station in Lagos that opened on March 21, 2026, giving African developers in-person support during The Synthesis.
"The infrastructure behind AI agents will determine whether their actions can be trusted," said Enoch Mbaebie, a spokesperson for Ethereum Nigeria, commenting on the Lagos hub.
Africa processes roughly 70 percent of the world's $1 trillion mobile money market, with 40 percent of Sub-Saharan African adults holding mobile money accounts as of 2024. DealForge's agent-to-agent contracting model, with its built-in escrow and audit trail, maps directly onto cross-border trade settlement, a persistent friction point across the continent.
In South Asia, India's large pool of AI engineers and Pakistan's active blockchain policy push, including a State Bank of Pakistan CBDC pilot targeted for mid-2026, make verifiable agent infrastructure relevant to builders and regulators alike. The MCP Dev Summit in Mumbai, scheduled for June 14 to 15, 2026, signals further regional momentum around these technical questions. EigenCloud's AgentKit is accessible globally, and both the Starknet and Base deployments are open to developers in Bangalore, Karachi, Dhaka, Colombo, and beyond.
The 2026 roadmap for EigenCompute includes moving beyond hardware-based trust toward zero-knowledge proof verification, which would remove the need to trust Intel's hardware guarantees at all. That shift is not yet delivered.
For now, EigenCloud's TEE-based approach represents a practical, if imperfect, bridge between autonomous AI systems and the auditability that financial applications require. The projects built during The Synthesis offer early evidence that this bridge can hold real weight. Whether builders adopt it at scale is the next question.