Platform Engineering - MuleSoft: A reengineering evolution in Salesforce

Building Flex Gateway

MuleSoft’s Anypoint team built the Flex Gateway, a high-performance, Envoy‑based API management layer, using open‑source technologies like Envoy, Fluentbit, WebAssembly (WASM), DataWeave, Go and Rust. There is a defined platform engineering happening here with a focus on developer experience, streamlined CI/CD integration and modular architecture.

According to Evangelina Martinez Ruiz Moreno, writing on the company’s own technical blog, “My MuleSoft team concentrates on the developer experience of API development and on the management of the API instances. We provide tools for writing specifications, testing APIs, and managing their access. This includes adding security constraints such as restricting access to certain clients or setting request limits.”

Her team offer this experience through the company’s own API platform. They develop the runtimes where the API or API Proxy is deployed and where policy enforcement takes place. She also notes that the platform engineering practice adopted and implemented a milestone-based approach to enhance the delivery process. 

“These milestones served as checkpoints for conducting internal demos and gathering valuable feedback from both internal teams and external stakeholders. This approach facilitated early integration among the various components involved in the execution,” she said.

MuleSoft’s platform engineering approach

We can dive deeper into MuleSoft’s platform engineering approach and highlight its work on core API gateway development. The company underlines the fact that Flex Gateway is built with open source technology and iterative user-focused design principles that also champion sustainable engineering practices that are baked into its runtime and application design practices. 

The company also evidences key platform engineering abilities in terms of automation and being what it likes to call “AI-first” in its development processes that use Agentforce to automate API documentation and metadata generation. There is also tight technical alignment with Salesforce’s platform services for API governance, developer tooling, and CI/CD pipelines.

Reengineering the Salesforce platform

According to the Salesforce Architecture blog, shifts in the market and industry in recent years necessitated a reevaluation of the Salesforce Platform. 

This involved a platform engineering level reimagining, refactoring reengineering process that started in 2020 in response to the emergence of public cloud providers who invest heavily in infrastructure, the increasing data residency and regulatory demands across different sectors and countries… and the need for handling real-time data and transactions at a much larger scale due to the rise of social and mobile technologies.

As noted above, MuleSoft has been a part of Salesforce for seven years, so platform engineering changes would naturally filter across both technology bases once applied. The changes needed to reflect advancements in machine learning and AI (particularly in generative – and now agentic – AI services) while also catering to growing requirements for cybersecurity, system availability, performance and resilience.

Loosely coupled, but coherent 

There was a demand from customers for an integrated suite that balances a loosely coupled but coherent architecture i.e. exactly what platform engineering embodies.

“In response to these challenges, Salesforce embarked on a mission four years ago to completely transform its platform from the ground up. This initiative aimed to address the aforementioned challenges and lay the groundwork for the next generation of applications and customer use cases, all while upholding our application availability goals,” said Srini Tallapragada, president and chief engineering and customer success officer at Salesforce.

Tallapragada reinforced the point by saying that the launch of Agentforce at Dreamforce 2024 represented the culmination of this extensive effort, involving thousands of Salesforce technology and product organisation team members. 

“The introduction of Hyperforce [a foundational infrastructure that has evolved from first-party data centers to public cloud providers, enhanced with Salesforce technologies for secure, compliant, highly-available hosting] Data Cloud, and generative AI technologies has significantly enhanced the platform’s capabilities, ensuring it remains at the forefront of innovation while maintaining trust and reliability. The successful migration of the majority of our customers to this new platform underscores the ingenuity and dedication of our engineers,” said Tallapragada.

NOTE: Salesforce Hyperforce has been described as a “complete re-architecture of the Salesforce platform” for the public cloud. Hyperforce infrastructure is composed of code rather than hardware, so that the Salesforce platform and applications can be reliably delivered to locations worldwide, which enables user choice and control over data residency.

MuleSoft integrated Agentforce (Salesforce’s digital labour platform) into its developer tooling to automatically generate API documentation. By combining Anypoint Exchange, CloudHub and the API Experience Hub with Agentforce, the company intends to automate repetitive engineering tasks that align with platform engineering principles.