USDA - AI-Assisted App Development - MRAS

The United States Department of Agriculture (USDA), through the Office of the Chief Information Officer (OCIO), is seeking to modernize its software development practices by integrating enterprise-grade AI-enabled coding solutions. The General Services Administration (GSA) is conducting market research to identify potential solutions that are FedRAMP-authorized and capable of securely scaling across USDA's diverse development environments. These efforts are part of a broader strategy to enhance IT strategy, cybersecurity, cloud adoption, and application modernization across USDA's varied mission areas, including farm services, nutrition assistance, rural development, food safety, and more.

The USDA OCIO supports a wide-ranging application development ecosystem that underpins critical programs and internal systems. Currently, the department lacks standardized access to AI-enabled coding tools that meet federal security requirements. This gap has led to slower prototyping cycles, increased developer workload for repetitive tasks, longer onboarding times for new technical staff, and missed opportunities for improving test coverage and documentation quality. To address these challenges, USDA seeks secure AI-enabled coding capabilities delivered through three primary modalities: IDE-Based AI Coding Assistance, CLI-Based Agentic Coding tools, and Browser-native AI application development environments.

IDE-Based AI Coding Assistance involves integrating AI tools directly into development environments to offer features such as code completion and generation, inline chat assistance, automated test generation, and code review. CLI-Based Agentic Coding focuses on terminal-based systems capable of multi-step code generation and refactoring along with support for continuous integration/continuous deployment (CI/CD). Browser-native platforms aim to enable rapid application creation from natural language prompts in a collaborative web environment. All these modalities must operate within USDA's security framework leveraging FedRAMP-authorized environments.

Potential solutions must meet specific attributes such as FedRAMP Authorization at Moderate or High levels or be in the process of obtaining such authorization by 2026. Additionally required are data protection measures ensuring customer data is not used for model training beyond processing necessity; enterprise administration features like Single Sign-On support; scalability across thousands of users; analytics capabilities; commercial maturity with proven enterprise deployments; and security measures producing secure code compliant with federal standards while mitigating risks like prompt injections or supply chain vulnerabilities.