Summary
At AI Engineer, Armanas Povilionis from Alithea Bio presents Froglet, an open-source protocol designed to give AI agents a verifiable transaction layer when collaborating across organizational boundaries. The core argument: adding more tools to agents improves local efficiency, but scientific and enterprise workflows require something fundamentally different — a way for agents to discover external services, negotiate execution terms, pay for work, and receive cryptographically signed receipts proving what was done and when.
Froglet positions itself as infrastructure between moving parts rather than a replacement for existing agent harnesses or payment rails. Each Froglet node generates a key pair for identity and signs every artifact in a chain, making the full execution record tamper-evident. Nodes can play multiple roles — provider, requester, or marketplace — using the same software. Integration is available as an MCP server, an OpenClaw plugin, or a NemoClaw plugin, with Claude as the primary intended human interface for both providers and requesters.
The live demo shows three scenarios: connecting to a hosted trial node at froglet.try.froglet.dev to generate a temporary identity token, running a local Docker-based node, and using Claude to interact with a local Froglet installation to configure a service. The talk frames the protocol as preparation for a near-future where agents manage their own budgets autonomously — discovering providers, ordering resources, and coordinating cross-organizational workflows in the same way an executive chef manages a supply chain rather than cooking alone.
📺 Source: AI Engineer · Published July 18, 2026
🏷️ Format: Keynote Launch
