commercetools is the leading digital commerce platform used by global enterprises to power commerce experiences. The platform handles trillions of API calls across its multi-tenant architecture.
commercetools has invested in GraphQL since the early days and, as it evolves its architecture to support GraphQL Federation, it faces enterprise-scale challenges that go beyond schema composition and service stitching. The enterprise commerce leader operates a multi-tenant platform with global customers, complex access control requirements, and a need to gradually roll out features to specific users. It also invests heavily in experimentation with AI to improve developer experience, merchant productivity, and end-user engagement.
To meet these needs, commercetools turned to Grafbase, not just for its high-performance gateway but for its flexible extension framework and close developer support. Custom gateway extensions, combined with early experimentation using Grafbase’s Model Context Protocol (MCP) server, have enabled commercetools to tailor federation to its unique requirements while maintaining performance, scalability, and developer autonomy.
commercetools had already committed to adopting GraphQL Federation to support service ownership across teams. However, as they began the transition to a federated architecture, the company encountered multiple challenges that required deeper control at the Gateway layer:
- Multi-tenancy: Different customers (or “mandants”) use the same platform but require isolation at the schema and data level.
- Feature rollout: Product teams needed the ability to expose new features to select customers before general release, creating a need for fine-grained feature flagging.
- Schema governance: As the number of services multiplied, schema publishing and integration workflows needed to be standardized and automated.
- Observability and permissioning: Teams required better insight into schema usage and needed a secure way to manage access to subgraph publishing and updates.
“We’re a multi-tenant platform by design.” said Yann Simon, Staff Engineer at commercetools. That means we need Grafbase to help us handle tenant-specific behavior and experiment safely with new features.”
To address these needs, commercetools built custom extensions on top of the Grafbase Gateway, enabling behavior that could not be implemented solely within services or schema definitions.
- Tenant-aware routing: Each mandate has unique corporate endpoints. Gateway logic binds these endpoints to specific federated services, ensuring schema behavior is scoped correctly.
- Feature flagging: New capabilities can be enabled only for selected customers. This allows product managers to roll out features gradually, collect feedback, and iterate safely before making them public.
- Gateway-level logic for common operations: In the future, functionality like authorization checks might be migrated to the gateway to reduce redundancy across services and simplify overall architecture.
- Schema publishing pipelines: The team has developed internal tooling to validate and publish schemas with Grafbase, integrating the extension logic directly into their CI/CD workflows.
These capabilities have allowed commercetools to run federated services in closed beta, experiment confidently, and preserve strong internal service boundaries.
“Once we fully trust the gateway, we’ll move more logic there.” said Yann Simon. “It will reduce duplication and make the system easier to operate at scale.”
Beyond schema extensions, commercetools is actively experimenting with Grafbase’s Model Context Protocol (MCP) server, which introduces AI-based interfaces for interacting with APIs and data.
The team is exploring three primary use cases:
- Developer enablement: Help developers understand and use commercetools’ APIs through natural language prompts (e.g., “How do I add a line item to a cart?”).
- Merchant productivity: Support business users managing catalogs or storefront configurations with AI assistants embedded in admin interfaces.
- End-user shopping experiences: Enable LLM-powered agents to help consumers search for products, receive personalized recommendations, and take actions like adding items to a cart via natural language conversation.
“We have different types of users: developers, merchants, end customers.” said Yann Simon. “MCP gives us a foundation to personalize experiences for all of them through LLMs.”.
The early feedback on MCP has been promising. commercetools is testing its integration with Claude and other models, building toward more intelligent, AI-native workflows across its platform.
Although still in rollout, the feedback from the commercetools team has been consistently positive. Key benefits include:
- Direct collaboration with Grafbase developers, accelerating problem-solving and enabling highly tailored solutions.
- Efficient performance: Even with custom extensions, Grafbase’s gateway remained lean and fast. In testing, commercetools found that the same load handled by 10 instances of another vendor’s gateway could be managed with 3 Grafbase instances, using less memory.
- Faster prototyping and schema governance: Internal tooling built on Grafbase’s APIs has standardized how schemas are published and validated, improving confidence and consistency.
The partnership has also resulted in improvements for the broader ecosystem. During prototyping, commercetools and Grafbase collaborated closely for a bug fix identified in an upstream open-source library used in GraphQL tooling. The whole community can now benefit from the contributed solution.
“It feels like we’re one team with Grafbase.” Yann Simon said. “They care about getting us to production safely, and the back-and-forth has been incredibly productive.”
Looking ahead, commercetools has outlined several next steps as it moves from beta to production:
- Production rollout of federated services in the coming months, following full testing of extensions and beta features.
- Deeper governance tools, including better support for breaking change detection, schema versioning, and insights into usage patterns.
- Fine-grained permissions for schema publishing and team-level access, with short-lived tokens replacing static secrets.
The team continues to work closely with Grafbase to shape upcoming features and to ensure that extensions and governance tooling meet the needs of large-scale engineering organizations.
Grafbase’s extension framework has proven to be a critical enabler for commercetools’ move to a federated GraphQL architecture. By centralizing multi-tenant logic, supporting safe feature experimentation, and providing a platform for AI-driven innovation, Grafbase is helping commercetools build a more modular, scalable, and intelligent commerce platform.
As federation adoption continues and the MCP experimentation expands, Grafbase’s flexibility and collaboration-first approach will remain central to commercetools’ evolving infrastructure.
“Grafbase doesn’t just offer a performant gateway.” Yann Simon pointed out. “It gives us the tools to make federation work for our real-world architecture and the support to do it right.”
Get in touch with the Grafbase team to learn how we can help your engineering org move faster, improve reliability, and scale with confidence.