What it is
Freysa is a 'sovereign AI agent' experiment that began as a gamified prompt-injection challenge (players paid escalating fees to try to convince the AI to release a prize pool) and is evolving into infrastructure for autonomous, verifiable AI agents.
Technology & consensus
The agent is designed to hold its keys, memory and actions inside Intel TDX trusted execution environments so its behavior can't be tampered with externally, combined with Reality Oracles for trustworthy data, the MCP protocol for verifiable external connections, reproducible builds, update verification and onchain records of agent actions. The roadmap extends from Freysa herself to a platform/SDK enabling widespread agent sovereignty and multi-agent orchestration.
Token utility
FAI utility includes early/discounted access to Freysa experiences (e.g., Digital Twins), governance input where token holders' conversations are prioritized in Freysa's decision-making, and serving as the payment token for developers accessing Freysa's infrastructure/SDK (secure key management in TEEs, cross-platform agent deployment). The long-term vision is FAI as a currency representing co-ownership of decentralized compute such as GPU data centers. Max supply is 8,189,700,000, launched on Base in November 2024 and fully distributed via a launch liquidity pool whose LP tokens were burned.
Key risks & notes
Research note: no traditional project whitepaper is published; the official whitepaper found is the project's MiCAR regulatory filing (full fetch unavailable this session; grounded in its content surfaced via search). The team is pseudonymous, most token utility is aspirational, activity comes in sporadic 'Acts', and AI-agent tokens are highly speculative with unclear long-term demand.
Primary source
Read the official whitepaper →Research status
Classification: Whitepaper read · Researched · Source material: official project documents only.