Alitheia bio's Froglet: Agents Need Verifiable Receipts

Armanas Povilionis, presenting from Alitheia bio, argues that the path to truly automated scientific research lies not in providing AI agents with more tools, but in establishing a framework for verifiable collaboration. After a decade in life sciences collaboration projects, Povilionis contends that current tool-centric approaches are insufficient. True scientific advancement, he posits, hinges on agents being able to collaborate autonomously, which requires a chain of verifiable receipts to ensure trust at scale.

The Michelin Star Restaurant Analogy

Povilionis uses the analogy of a Michelin-star restaurant to illustrate the complexity of scientific work. Simply equipping a chef with better knives or more pans (analogous to more tools for agents) only enhances individual, local efforts. A high-quality restaurant, however, depends on the entire supply chain: reliable suppliers, consistent service, and the ability to reproduce outcomes reliably. This complexity mirrors the distributed nature of data and specialized algorithms across organizations, often trapped in silos.

Froglet: The protocol for agentic transactions

Alitheia bio's vision for Froglet is to address this challenge. As agentic workflow automation matures, organizations will not just allocate token budgets but will empower agents to manage their own budgets for discovering services, requesting data, negotiating execution, and handling cross-organizational payments. In this scenario, agents evolve from mere cooks with better tools to executive chefs, managing the entire process from sourcing ingredients to coordinating operations and maintaining meticulous records.

Froglet is designed as the protocol for agents to discover, transact with, and receive verifiable receipts for external data and services. It operates between various components, integrating with different payment rails, agent harnesses, execution environments, and network transport protocols. Crucially, Froglet does not mandate a uniform software stack; it requires only a common interface.

From Bespoke Projects to Minutes-Long Workflows

The current reality for close scientific collaboration often involves bespoke enterprise projects that can take years and cost millions before a single reusable workflow emerges. Froglet aims to drastically simplify this process. Once an organization deems a resource shareable, a provider can expose it via Froglet. An agent can then discover it, understand its terms, request the work, and receive a verifiable receipt, a process that can cost a few thousand tokens and take mere minutes.

Verifiable Receipts and Trust

The Froglet network consists of homogeneous nodes, where each actor plays a role as a requester, provider, or a specialized marketplace node. The protocol solves the critical problems of how requesters find providers, how trust is established, how payments are settled, and how execution is proven. Every interaction generates signed artifacts, creating a chain of evidence, from descriptors and offers to quotes, deals, invoices, and final receipts. This chain is tamper-evident; any alteration breaks the chain.

Payments within Froglet can include a base payment to protect providers from denial-of-service attacks and a success fee to protect requesters from malicious providers. Povilionis emphasizes that Froglet itself is not an AI agent but a foundational interface enabling agents to discover each other and execute deals directly across organizational boundaries. This, he concludes, unlocks the potential for autonomous scientific progress.