When Fetch.ai began life as a research project at Cambridge, artificial intelligence was still the domain of centralised tech giants. Today, it sits at the heart of a decentralised alliance reshaping the future of privacy-preserving AI agents.
“We started off as an advanced research project at Cambridge,” said Nikolay Dimitrov, DevRel and co-marketing lead for Fetch.ai and the Artificial Superintelligence (ASI) Alliance. “By 2021, we had launched our mainnet using the Cosmos SDK, and by late 2022 we introduced the micro-agent framework, our most mature framework yet for building autonomous AI agents.”
Fetch.ai’s framework allows developers to deploy AI Agents that represent people, places or products. These agents can complete complex tasks: from booking logistics to participating in marketplaces, coordinating transport, or acting as digital assistants. At the core lies an agent-to-agent communication protocol, soon to be backed by a chain of trust and selective disclosure, allowing for verifiable yet private AI interactions, provided through a partnership with Secret Network.
In its own words, Fetch.ai provides an open infrastructure for building smart, autonomous services. Using Fetch.ai technology it is possible for the technically curious to deploy AI Agents (bits of code) to create intelligent connections between each other and real world systems and devices. Fetch.ai tooling is fully composable to allow for freedom of exploration, from simple AI-driven coordination tasks to complex business logic
In early 2023, Fetch.ai launched the Agentverse, a hosting and Integrated Development Environment (IDE) platform for building, debugging and deploying agents. By 2025, it had released ASI Mini, its own large language model.
“It’s not just a chatbot. We designed it to facilitate real-world transactions and human intentions,” said Dimitrov. “We complement this by building the Agentverse as a port city for agents.”
The ASI Umbrella Movement
Fetch.ai is now a founding member of the Artificial Superintelligence Alliance, a collective effort between four core projects: Fetch.ai, SingularityNET, Ocean Protocol, and Cudos. Their goal is to provide a unified, distributed AI tech stack and to develop AGI (artificial general intelligence) using a common token, currently FET.
“You’ve got Cudos for compute, Ocean for data lakes, SingularityNET for research, and Fetch.ai for applications,” said Dimitrov. “Together, we’re blasting through bottlenecks in AI development. It’s not just a merger, it’s a movement.”
The ASI Alliance enables developers to build AI models that can reason, act, and communicate securely across networks. Fetch.ai’s contribution is its mature agent’s framework with products such as Agentverse.ai, asi1.ai and the ASI Wallet. In addition, Fetch.ai’s agentic-LLM called ASI:One now will now be bolstered by its partnership with Secret Network.
Privacy Meets Autonomy: The Secret Partnership
In 2025, Fetch.ai integrated its AI agent framework with Secret Network’s Confidential Virtual Machine (CVM), unlocking two critical features: privacy-preserving agent interactions and verifiability of agent behaviour without actually needing to see the code..
“This is about foundational value unlocks,” said Dimitrov. “Secret’s secure enclave-based technology allows inference to be protected from third parties, even hardware owners. You get privacy without breaking usability.”
Inference is the process whereby the actual computation where an AI model processes sensitive data to produce an output (like diagnosing a condition, summarizing legal text, etc.) is done inside a confidential environment and nobody else can see what’s happening, not even the owner of the hardware it’s running on.
The partnership enables:
- Confidential agent communication: Agents can interact without exposing sensitive data, leveraging Secret’s Trusted Execution Environments.
- Verifiable execution: Developers and users can ensure that an agent’s code has not been tampered with, a critical requirement for enterprise and regulatory trust.
“Verifiability and privacy at the infrastructure level reduce the lift for developers,” said Dimitrov. “You don’t have to bake in complex encryption yourself. You just inherit trust.”
Use Cases: Legal, Medical, Industrial
This architecture isn’t theoretical. It will be applied in high-stakes sectors where selective disclosure and verifiability are non-negotiable. Here are a number of use cases being explored.
Legal
Class action lawsuits often involve gigabytes of sensitive discovery documents. “Large language models are perfect for analysing that data,” said Dimitrov, “but lawyers can’t risk exposing it. Now they don’t have to. You get the efficiency of AI without compromising privilege.”
Medical
Fetch.ai has released multiple specialised AI models in the healthcare space, including tools for breast and prostate cancer detection. These models are deployed within the Agentverse and integrated with Secret’s infrastructure.
“We started here because healthcare has the most urgent need for privacy,” said Dimitrov. “Imagine asking a general practitioner AI model for a second opinion, knowing your data stays private. That’s game-changing.”
Industrial
Agents can also represent trucks, ports, parking lots and terminals. These machine-to-machine interactions can automate logistics and optimise transport infrastructure.
“A truck can book its own parking or schedule cleaning for its container,” said Dimitrov. “That data needs to be private and the agents need to be verifiable, especially when money and compliance are involved.”
The Chain of Trust
What underpins all of this is what Dimitrov calls a “chain of trust.” From the agent code to the data flows to the execution environments, each layer can be verified.
“In the physical world, trust is inferred, does the restaurant smell nice? Is the staff polite? In the digital world, trust isn’t inferable. It must be established. And this partnership allows us to do that.”
The goal isn’t just to win business. It’s to raise the ethical bar and add impact.
The Vision of Fetch.ai
Fetch.ai doesn’t want to build tools for surveillance or ad targeting. Its vision is to empower users, not extract from them.
“Too much of AI’s history is spyware and adware,” said Dimitrov. “Early applications were built to monetise human behaviour. We want to turn that around.”
That means giving users their own agents, their own digital assistants, their own Jarvis, like Tony Stark’s AI butler in Iron Man.
“Wouldn’t you want your Jarvis to be private?” he asked. “Not owned by a corporate that might weaponise your data, but something verifiably yours.”
Fetch’s roadmap is built around that idea: open tools, agent marketplaces, AI models that support professionals and enterprises, and a privacy-first ethos embedded at every layer.
“Every sufficiently high-impact use case eventually requires privacy,” Dimitrov concluded. “That’s not a constraint, that’s an unlock.”
With the ASI Alliance pushing forward and deep integration with privacy-preserving infrastructure like Secret Network, Fetch.ai is poised to deliver on the promise of autonomous intelligence: private, verifiable, and truly useful.
