Huawei unveils A2A‑T open‑source project at MWC Barcelona 2026, expanding agent

Huawei’s A2A‑T platform, unveiled on the main stage of MWC Barcelona 2026, is positioned as an open‑source framework that standardises agent‑to‑agent (A2A) communication across heterogeneous AI ecosystems. According to the official MWC report, the project supplies a unified API layer, a set of protocol adapters and a sandboxed runtime that enable third‑party bots, virtual assistants and edge devices to exchange intents, context and state without proprietary translation layers. The company said the codebase will be hosted on GitHub under an Apache 2.0 licence, inviting contributions from academia, startups and rival vendors. By decoupling the messaging format from the underlying inference engine, Huawei aims to lower integration costs for enterprises that currently stitch together disparate AI services.

The technical brief released at the show highlights three core components: the A2A‑T Core, which implements a lightweight publish/subscribe bus; the Agent Bridge, a collection of language‑agnostic adapters for Python, Java, Rust and Go; and the Compliance Suite, which validates message schemas against the emerging ISO/IEC 30170 standard for autonomous agents. Huawei’s engineering lead, quoted in the report, emphasized that the framework supports both synchronous RPC calls and asynchronous event streams, allowing real‑time coordination in scenarios such as smart‑factory orchestration, autonomous vehicle fleets and cross‑platform digital assistants. The open‑source model also includes a certification program that will label agents as “A2A‑T‑Compliant,” a move intended to foster a marketplace of interoperable services.

Industry observers noted that the timing of the announcement diverges from the hardware‑centric expectations that typically dominate MWC showcases. ZDNet’s live‑update coverage described the A2A‑T reveal as “the most surprising development of the week,” pointing out that Huawei has been investing heavily in AI software stacks since its 2023 Ascend AI‑Core launch. The outlet added that the open‑source project could accelerate adoption of Huawei’s own Ascend chips by providing a ready‑made software layer that abstracts hardware specifics, thereby enticing developers who might otherwise gravitate toward competing ecosystems such as NVIDIA’s CUDA or Google’s TensorFlow.

The Daily Mail’s brief on the event focused on the broader strategic implications, noting that Huawei’s push into open‑source software aligns with its recent efforts to diversify revenue streams beyond handset sales. While the newspaper’s coverage was limited to a single paragraph, it referenced the A2A‑T platform as part of a “larger push to capture the AI‑agent market” and suggested that the move could help the Chinese giant mitigate the impact of ongoing export restrictions on its semiconductor supply chain. The article also mentioned that Huawei simultaneously announced a new MateBook hybrid laptop, but did not link the two product lines directly.

Finally, 9to5Google’s “evleaks” feed, which aggregates unofficial disclosures from the show floor, reported that several early adopters—including a European telecom operator and a Japanese robotics firm—have already begun testing the A2A‑T runtime on pilot projects. According to the leak, the telecom partner is using the framework to enable dynamic hand‑off between voice‑assistant services across 5G edge nodes, while the robotics company is leveraging the Agent Bridge to synchronize swarm behaviours among warehouse drones. Both test cases underscore the platform’s promise of cross‑domain interoperability, a claim that Huawei reiterated in its press materials. If the open‑source model gains traction, A2A‑T could become a de‑facto lingua franca for AI agents, reshaping how vendors compete on the software layer rather than on proprietary hardware alone.