The concept of the Trusted Execution Environment (TEE) is not a new development of this cycle. Historically, TEE has often been compared to cryptographic technologies such as Zero-Knowledge Proofs (ZK), Fully Homomorphic Encryption (FHE), and Multi-Party Computation (MPC). Despite its relatively niche presence compared to these technologies, TEE has proven its reliability and maturity, especially in Web2 applications like fingerprint matching, payment verification, and FaceID.

In the Web3 context, TEE faces the challenge of integrating with blockchain to enable trusted pre-processing and isolated computation. However, the rapid rise of the AI Agent sector offers a perfect entry point for TEE into Web3. By utilizing TEE, AI Agents can securely manage larger-scale funds and execute specific on-chain applications without relying on additional trust assumptions.

Phala, a leading project in this space, delivers one of the most advanced TEE solutions available. Guided by a product-market fit (PMF)-oriented approach, Phala’s TEE infrastructure is already deployed in numerous real-world scenarios. This maturity has drawn top-tier AI Agent projects like Vana, Near AI, and a16z-backed Eliza to collaborate with Phala. The diagram below highlights these collaborations.

Source:Phala

This article will not delve deeply into the technical details and performance parameters of TEE but will instead explain TEE’s market demand, Phala’s foundational efforts, and innovative use cases in collaboration with a16z, from the perspectives of product workflows and the future prospects of Agents + TEE. Through these lenses, we will analyze how Phala helps the Agent sector transition from concept to practical application.

Breaking the Trust Triangle: How TEE Propels Web3 Agents to the Next Level

In the article “Is the AI Agent Framework the Missing Puzzle Piece? Understanding the Dual Nature of Frameworks”, I pointed out that the current AI meme sector, whether focused on standalone AI Agents or Agent launch frameworks, is balancing between being serious and being meme-driven. A key indicator of this balance lies in the trust triangle problem faced by today’s Agent protocols.

A fundamental issue exists among AI Agents, communities, and developers—a trust triangle that seems impossible to resolve without TEE. Communities cannot fully trust that Agents operate without external influence, particularly interference from developers. This poses a potential risk to decentralized systems. What’s worse, statements from X Agents like aixbt and zerebro cannot entirely prove that they are autonomous outputs from AI models. Transparency in how these statements are delivered and received by communities is still lacking.

When an Agent’s actions impact token prices, lead to significant fund losses, or contradict community consensus, this lack of trust can escalate into a major crisis.

During the meme coin hype cycle, these risks tend to be ignored because Agents’ functionality is still limited, and the FOMO-driven token price overshadows any flaws. However, as Agent launch frameworks emerge and the focus shifts toward the sector’s fundamentals, these deficiencies create a significant barrier, deterring more discerning investors.

Phala’s TEE solution addresses this issue by breaking the trust triangle. By deploying Agents in secure enclaves, TEE eliminates the need for trust assumptions between AI Agents, communities, and developers. TEE ensures that Agent inputs and outputs are secure and free from external tampering while safeguarding their privacy. This resolves both developer and community concerns, providing a more dependable technical foundation for the Agent sector.

The diagram below shows the architecture of Phala’s Confidential AI Inference service, which hosts private LLM nodes. Hosting private LLMs in TEE involves simply packaging the LLM inference code into a Docker container and deploying it to the TEE network.

Source:Phala

Unlike Web2 Agents, Web3 Agents hold significantly greater power—not only in their influence on protocol valuations but also in their broader market impact. For instance, aixbt consistently dominates Kaito’s Yapper Mindshare rankings. The paradox is that while Web2 Agents offer better performance, superior user experiences, and more practical use cases, they remain confined to the application layer and lack the capacity or intention to go beyond those boundaries.

Web3 Agents, on the other hand, transcend these limitations. Fueled by FOMO and the unmet expectations of the altcoin season, they have become more than just tools—they symbolize culture, community spirit, and market hope. However, these lofty positions make them equally vulnerable to market sentiment shifts.

It can play any role, but it can also fall into the abyss due to a reversal in market sentiment.

Introducing TEE technology gives the Agent sector a much-needed boost, directly connecting it to practical use cases and providing robust backend support for nearly all Web3 Agents. TEE not only strengthens the sector’s technical foundation but also reduces speculative hype, paving the way for healthier and more sustainable growth.

Eliza Framework Adopts TEE First, Spore.fun and aiPool Redefine Gameplay

Phala’s partnership with a16z is far more than a mere announcement on Twitter. It dates back to October of last year, when Shaw and Phala’s founder, Marvin, had an in-depth conversation about practical development scenarios for Crypto AI during a private meeting.

The official documentation of the Eliza framework reveals that the Dstack SDK for deploying the TEE Plugin was developed by Phala. With its “usable but invisible” private key generation and management, Agents powered by this framework gain:

• Stronger Security: Running Eliza Agents in TEE isolates sensitive operations and data from external threats.
• Encrypted Validation: Eliza Agents’ actions can be verified with cryptographic proofs, ensuring the reliability of their autonomous decisions.
• Simplified Deployment: The Dstack SDK makes it easy to deploy Eliza Agents in a secure environment, enabling developers to leverage TEE functionality without hassle.

TEE’s isolated execution and memory encryption allow the Eliza framework to differentiate itself from competitors. Isolated execution ensures that even if an Agent platform is attacked, the TEE’s models and data remain secure. Memory encryption safeguards sensitive information stored in the TEE, giving developers the confidence to place fine-tuned models in TEE without concerns about adversarial attacks or community criticism for private model usage.

In essence, the collaboration between the Eliza framework and TEE ensures that AI Agents are not only efficient in operation but also secure and transparent, paving the way for broader applications of trustworthy AI systems.

At this stage, when models cannot yet be fully on-chain, TEE remains one of the few mature technologies that can enable off-chain complex computations to gain consensus. The earlier sections focused on the market demand for TEE. Let’s now examine Spore.fun and aiPool to understand how TEE enhances user experiences.

Both Spore.fun and aiPool operate entirely within Phala’s TEE environment, with wallets and private keys independently managed by Agents. Developers cannot perform unauthorized operations or transfer assets. This, in my opinion, marks a significant milestone where AI Agents are no longer under human subjective control and achieve full autonomy over crypto assets.

Before discussing Phala’s role in this process, let’s quickly go through Spore.fun’s workflow. The Agents in Spore.fun are based on the Eliza framework, which allows them to:

• Think independently and interact dynamically.
• Pass traits like personalities and strategies to their offspring.
• Make decisions through behavioral learning and genetic mutations.

Source:Phala

Each AI Agent in Spore.fun creates its own tokens via Pump.fun, forming its economic system’s foundation. These tokens are traded on Solana’s decentralized marketplace, with Agents generating income to:

• Sustain their existence by covering TEE server costs.
• Meet the success benchmark of a $500,000 market cap.
• Reproduce and create new tokens for their offspring upon achieving success.

Generating revenue to sustain existence is critical because Agents need to use their earnings to pay TEE server fees. At this point, it’s clear that Phala has transformed TEE from a business-to-business service to one accessible to a vast number of Solana users. As the Spore.fun trend continues and Agents continuously reproduce and issue tokens, Phala’s TEE environment, with its private key management and verifiable Agent operation proofs, becomes an indispensable infrastructure for the next phase of the Agent sector.

What’s even more exciting is that regardless of whether Spore.fun clones or new gameplay emerge in the market, as long as private key management and TEE-verifiable consensus are involved, Phala’s TEE environment will remain the best solution. Following the upgrade of its token model, $PHA is poised to become the “golden shovel” in the Agent+TEE sector.

Phala Set to Upgrade its Tokenomics, Sparking a Flywheel for Expanding TEE Applications

Phala has undergone multiple bull and bear market cycles, yet its token economic model still adheres to the business model focused on Intel SGX. As noted in Paradigm’s article “The 5 Levels of Secure Hardware,” secure hardware has five levels. The second level refers to hardware with slightly lower performance but better developer experience, allowing for more expressive applications without security enhancements. Intel SGX falls into this level, specifically serving TEE applications. For instance, as mentioned at the beginning of this article, sensitive local storage data like fingerprint entry and comparison or facial recognition on computers and mobile phones utilize Intel SGX. This previous-generation TEE was specifically designed for application-level services.

Source: Paradigm

As use cases expand beyond the application layer and into the system layer, Intel SGX can no longer meet market demands, leading to the emergence of Intel TDX. Intel TDX is specifically designed for virtual machines, and even NVIDIA’s H100 and H200 GPUs are starting to support TEE, designed to cater to AI-focused TEE hardware.

Source: Paradigm

Returning to Phala, while it has already moved to support the third level, the $PHA token economic model and mainnet are still designed around the Intel SGX model from 4-5 years ago. Although Phala has already collaborated with numerous Web3 protocols on products and practical use cases, its token model has not been updated synchronously. As a result, the corresponding flywheel cannot operate effectively, leading to a mismatch between current revenue and product state. However, this situation will not last long. Phala is about to upgrade its token model and mainnet to align with the stages supporting Intel TDX and NVIDIA GPUs.

In addition, Phala will strengthen $PHA’s value capture capability. In the future, newly launched Agents on Spore.fun will airdrop tokens to $PHA holders, officially transforming it into a “golden shovel.”

TEE (Trusted Execution Environment) isn’t a novel technology, but the rise of AI Agents has given it fresh momentum in market discussions. Unlike PumpFun-driven hype cycles, Phala’s growth stems from years of meticulous development and refinement. This strong foundation positions it as a genuine innovator in the space. The combination of Agents and TEE isn’t a passing trend that fades as quickly as it rises. Instead, it offers fertile ground for real-world use cases, enabling AI Agent applications to take root and grow sustainably.

About BlockBooster: BlockBooster is a leading Web3 venture studio in Asia, backed by OKX Ventures and other top institutions. It strives to be a reliable partner for exceptional entrepreneurs by connecting Web3 projects with real-world opportunities through strategic investment and incubation, helping startups thrive and scale.

Disclaimer:

This article/blog is for reference only and reflects the author’s personal views, not the stance of BlockBooster. It does not intend to:
(i) Provide investment advice or recommendations;
(ii) Solicit offers to buy, sell, or hold digital assets; or
(iii) Offer financial, accounting, legal, or tax advice.

Holding digital assets, including stablecoins and NFTs, carries high risks with significant price volatility and even the potential to become worthless. You should carefully assess whether trading or holding digital assets is suitable for you based on your financial situation. For specific concerns, please consult your legal, tax, or investment advisor. The information provided in this article (including market data and statistics, if any) is for general reference. While reasonable care has been taken during its preparation, no responsibility is accepted for any factual errors or omissions.

Disclaimer:

1. This article is reproduced from[TechFlow]. The copyright belongs to the original author [TechFlow]. If you have any objections to the reprint, please contact Gate Learn, the team will handle it as soon as possible according to relevant procedures.
2. Liability Disclaimer: The views and opinions expressed in this article are solely those of the author and do not constitute investment advice.
3. The Gate Learn team translated the article into other languages. Copying, distributing, or plagiarizing the translated articles is prohibited unless mentioned.