Reported impact
Business outcome
Fintopia converted AI skills from loosely managed scripts into a measurable asset base. The company can now route developers toward reuse, reduce repetitive setup work, publish faster, and use real telemetry to decide which skills need optimization. It also established a governed layer spanning more than 50 models, more than 90 MCP integrations, more than 400 reusable skills, and more than 100 packaged plugins.
- Productivity Improvement
- Scalability
- Speed to Market
Solution
Fintopia built a SkillOps control plane on top of Alibaba Cloud AgentLoop to turn scattered AI skills into governed enterprise assets. Developers discover reuse candidates through a skill marketplace, scaffold new skills from captured context, and publish directly from Git into a managed distribution flow instead of passing scripts around informally. The runtime spans a layered internal AI stack: a model plaza that routes across more than 50 models, an MCP plaza that standardizes connections to more than 90 internal systems and data platforms, a skill plaza that hosts more than 400 reusable business skills, and a plugin plaza that packages them for cross-team distribution. AgentLoop and its LoongSuite collector then capture real usage through local hooks, remove PII, attach metadata such as repo and branch for coding skills, and feed evaluation, regression, and release-gate loops so Fintopia can improve AI behavior with evidence instead of intuition.
Data flow
Employees and developers invoke skills through Fintopia's internal plazas, while AgentLoop captures traces, usage outcomes, and skill-specific metadata through LoongSuite. The collected data is desensitized, tagged back to the owning skill, scored by evaluators, and then fed into release gates and iteration loops before updated skills are republished.
