When and How to Adopt Micro‑Frontend Architecture: Insights from Ant Group’s Liu Kui

🙋🏻‍♀️ What scenarios suit micro‑frontend, how to judge the best timing, how to split business most efficiently, and whether the trend has passed? This interview with Ant Group frontend engineer Liu Kui explores his micro‑frontend practice and thoughts.

Recent rapid internet growth has increased business complexity and maintenance costs. Micro‑frontend emerged as a hot architectural approach, with solutions like Ant Group’s qiankun and JD’s MicroApp. However, practical adoption reveals many challenges.

Micro‑frontend is an architectural concept similar to micro‑services, applying the micro‑service mindset to the browser by aggregating multiple small front‑end apps into a single application while allowing independent development and deployment.

Liu Kui, the author of qiankun and a contributor on GitHub, discusses his work at Ant Group’s Experience Technology Department. He explains that their micro‑frontend R&D platform not only builds foundational frameworks and libraries but also designs product‑level processes covering the entire lifecycle of micro‑applications—from creation on internal platforms, relationship description, gray‑scale release, to monitoring and incident response.

He also describes extending this dynamic mechanism to other middle‑platform scenarios, such as page hosting, gray‑scale deployment, and multi‑environment deployment (public cloud, private cloud), which are common challenges regardless of micro‑frontend adoption.

Key challenges Liu identifies are:

Determining the granularity and boundaries when splitting a monolithic application.

Managing dependency reuse among multiple micro‑applications.

Governing micro‑applications at scale.

For splitting, page‑level granularity works well, but extracting UI fragments as independent micro‑apps can be tricky. He suggests evaluating service completeness rather than just UI reusability, using the rule: if a micro‑app frequently communicates with others, the split may be too fine.

Regarding dependency reuse, Liu notes that shared dependencies can introduce hidden coupling, contradicting micro‑frontend principles. While bundless/ESM techniques can mitigate this, a mature solution is still under exploration.

Governance issues arise when many micro‑apps interact, leading to version mismatches or uneven availability. Early platform‑side data collection and systematic product design can help address these problems.

He outlines when micro‑frontend is appropriate: large, continuously delivered applications, multiple teams, and integration needs. If none of these apply, micro‑frontend may be unnecessary.

Typical usage involves routing to embed other apps, but finer‑grained integration—such as embedding a micro‑app in a drawer or card within a DevOps or monitoring platform—is also emerging. Liu emphasizes the need for efficient production and consumption of micro‑apps, proposing two approaches: dedicated directories for shared services or marking UI blocks for automatic extraction during static builds.

Despite perceived waning hype, Google Trends shows steady interest, and Liu observes increasing micro‑frontend mentions in job postings, indicating broader adoption.

He believes the foundational technologies are stable, but the next step is standardizing the entire micro‑app lifecycle—build, deployment, consumption, and governance—potentially through a community‑driven whitepaper.

Regarding qiankun, Liu announces the upcoming 3.0 version, which will modularize capabilities into separate packages (loader, sandbox, etc.), adopt the new ShadowRealm API for sandboxing, and provide UI‑framework‑specific components.

Finally, Liu encourages engineers to engage with open source, starting with small contributions to projects they already use, as a path to personal growth and company value.