What tech stack does Brex use?

Brex runs a Kotlin-first microservices architecture, having publicly migrated from Elixir starting in 2022. The migration rationale — documented on the Brex engineering blog — cited static typing (which Elixir lacks), better gRPC tooling in Kotlin, and JVM ecosystem maturity for financial applications. Golang is used for select performance-critical services alongside Kotlin. The 2022 transition was a significant undertaking for a company of Brex's scale and reflects the engineering culture's willingness to make painful platform bets for long-term technical health.

All production workloads run on AWS: EC2 for compute, EKS for Kubernetes orchestration, ECR for container images, RDS for managed PostgreSQL, S3 for object storage (receipts, statements, audit logs), and CloudFront for content delivery. Kafka handles high-volume event streaming between microservices — critical for a platform processing billions of dollars in card transactions. Bazel is used for build tooling, enabling fast incremental builds across a large polyglot monorepo. ArangoDB, a multi-model graph database, handles specific graph-structured data workloads — likely authorization graphs or relationship models.

CTO James Reggio's 'Brex Hacker House' (2024–2025) layered a significant AI agent capability onto this foundation — building LLM-powered compliance automation, policy enforcement, and financial insights directly into Brex Empower. The AI layer is built in-house rather than relying on a third-party platform, consistent with Brex's build-oriented engineering culture.

The Brex web application is built with React and TypeScript, using Federated GraphQL as the API layer between frontend clients and backend microservices. Federated GraphQL enables independent teams to own their subgraphs while the frontend composes a unified data graph — a modern pattern that scales well at Brex's engineering org size. Retool powers internal admin and operations tooling, a common pattern at fintech companies where operational surfaces need to be built quickly.

For analytics and data warehousing, Brex uses Snowflake as the central analytical data warehouse, with Airflow for pipeline orchestration and dbt likely used for data transformation (a ubiquitous pairing with Snowflake at this scale). Brex's own public documentation confirms the Snowflake migration for data science use cases including fraud detection, underwriting, and product analytics. PostgreSQL via AWS RDS handles transactional data at the service level.

On the GTM and marketing side: Segment is the customer data platform for event collection and identity resolution; Amplitude handles product analytics and experimentation; Marketo manages marketing automation and email campaigns; Optimizely and Hotjar cover A/B testing and session recording. These tools are detected via Himalayas, BuiltWith, and Blind discussions, and are consistent with a B2B SaaS company at Brex's scale and enterprise-sales motion.

Brex's Kotlin/AWS/Kubernetes stack means they are deep AWS customers — any infrastructure, security, observability, or developer-tooling vendor that integrates natively with AWS and Kubernetes is already speaking Brex's language. Vendors requiring GCP or Azure-specific deployment will face meaningful friction. The Bazel-based monorepo build system signals that Brex takes build performance and reproducibility seriously — developer experience tooling that integrates with Bazel is a natural conversation.

The Snowflake + Airflow + dbt data stack signals strong appetite for data products compatible with Snowflake — data quality, governance, reverse-ETL, or data observability tools should lead with Snowflake compatibility. Segment as the CDP means any identity-resolution or enrichment tool needs to play well with Segment's event schema. Brex's in-house Retool usage for operational tooling suggests a high build-vs-buy threshold for internal surfaces — you'll win more by solving problems Retool can't (compliance automation, financial controls, ERP-grade fidelity) than by competing with it.

Post-Capital One acquisition, Brex's technology decisions will increasingly intersect with Capital One's existing vendor relationships and enterprise security requirements. Vendors already on the Capital One approved-vendor list have an advantage in Brex deals; new vendors will face Capital One's security review process as well as Brex's own. The Elixir-to-Kotlin migration also signals that Brex is willing to make large-scale platform transitions for long-term technical health — a culture that may prefer to build point solutions in-house rather than buy commodity tooling.