Key Highlights
- Tether’s new assistant shows that AI automation can run privately on basic consumer hardware without cloud dependency.
- The system uses the Model Context Protocol (MCP) to bridge local reasoning with professional productivity tools like Asana.
- An upcoming open-source release aims to decentralize AI access and provide a transparent alternative to centralized corporate models.
Tether CEO Paolo Ardoino revealed the results of live tests conducted on the company’s newly developed QVAC AI Assistant. The development marks Tether’s entry into the decentralized AI space, focusing on privacy and hardware accessibility rather than centralized cloud computing.
In an X post on Thursday, Ardoino showed the assistant working with the Model Context Protocol (MCP) to manage professional activities and functions, including task creation in Asana. The assistant has the capability to perform complex administrative operations through local inferences on consumer-grade computer devices.
The demo highlighted the AI’s ability to carry out multi-step operations with a sub-average laptop GPU. This proved that high-level reasoning does not need large server farms.
During the test, the QVAC interface processed a request for the organization of a project release. As a result, the system generated a main task and a subtask using the Asana productivity tool.
High-performance metrics
The system’s internal thinking process was reflected through the model thinking toggle buttons before the tool call process. The metrics reflected through the terminal for the efficiency of the model included a time of 1062.1 ms to generate the first token and a speed of 34.6 tokens per second.
Over the past year, Tether has invested more in telecommunications, energy production, and artificial intelligence infrastructure. Ardoino has previously argued for the need for independent, peer-to-peer technology that lessens reliance on big tech companies. Tether first launched QVAC AI in May 2025 and further expanded the QVAC Genesis II AI Dataset to 148 billion tokens in December 2025.
By creating an assistant that operates entirely on a user’s local machine, Tether aims to address rising concerns about data privacy and the risk of censorship with centralized AI models.
QVAC as open source
As for the future potential of the technology, it will likely focus on making AI agents more accessible. Ardoino confirmed that the QVAC AI Assistant will become open-source.
Developers will have the opportunity to examine the code and implement custom solutions by using the model context protocol. Assuming that the performance of the tool is sustained on low-end hardware, it might help to make it more accessible to individuals in areas where internet connection is scarce or to those who cannot afford high-end computing hardware.
It operates via the MCP, meaning it might support other third-party applications other than Asana. In turn, it can serve as a localized alternative to mainstream AI agents in enterprise environments.
Independent technology development
The QVAC AI Assistant testing is a huge step for Tether, which is changing from a financial services provider to a developer of independent technology. The company was able to develop a working prototype of local and independent AI with 100% local inference and reasoning on a standard laptop.
The upcoming open-source release will be the ultimate test of whether this tool can attract interest from the broader developer community and provide a feasible path for decentralized digital assistants.
Also Read: Tether Introduces Searchable Directory for USD₮ Services
