Warp on Tuesday launched Warp Factories, open infrastructure designed to help developers build cloud software factories—agentic systems that automate tasks across the software development lifecycle. The agent development platform describes Warp Factories as "the building blocks" for creating scalable factories, positioning the infrastructure as a fix for two common engineering pain points: measuring and boosting coding agent return on investment, and maintaining governance and control. Warp founder and CEO Zach Lloyd predicts software factories will become as widespread as CI/CD within the next few years.
The infrastructure includes features built to boost coding agent ROI over time through evaluations and benchmarks that measure effectiveness, plus integrated self-improvement capabilities and memory. Developers can access queryable metrics covering agent throughput, cost, quality, and ROI through the Factory control room, API, and Factory MCP. Scorers assess how work items progress through the factory, tracking elements like token expenditure, code quality, and whether work introduced bugs. Those scores then drive self-improvement loops and benchmarks, with "observer" agents scoring selected agent runs and hunting for improvements by tweaking variables like the harness, model, or context before submitting pull requests to enhance underlying factory functionality. On governance, the infrastructure offers version-controlled factory definitions, distinct agent definitions, plus skills, MCPs, and permissions.
According to Lloyd in a blog post, many organizations already have engineering teams building cloud software factories in-house, but he argues that's too large a project for internal development. Warp Factories provides the core components to accelerate development without forcing organizations to give up flexibility, programmability, customization, or ownership—what Lloyd calls making sure "the annoying bits [are] taken care of." Lee Faus, founder and CEO of Atomic Software and former global field CTO at GitLab, tells The New Stack he considers the software factory "inevitable," though he's more cautious about the timeline. "But before software factories become as foundational as CI/CD, the industry needs to solve a deeper infrastructure problem," Faus says, specifically highlighting the importance of tracing agent work and establishing a system of record for the factory.
Industry experts see software factories gaining momentum but expect a slower rollout than Warp's founder anticipates. Erik Gfesser, a longtime engineer, tells The New Stack he expects software factory adoption will likely fragment across multiple vendors, similar to the early stages of CI/CD. He points out that while the Warp client is open source, the server, the Warp Drive backend, and OZ—Warp's agent orchestration layer—are proprietary, with OpenAI named as the founding sponsor of Warp's open source repository. Faus adds that governance extends beyond controlling agent operations to proving what agents actually did: "A software factory without a record of change risks becoming a very efficient way to manufacture code that nobody can fully explain." As software factories accelerate code generation, the harder challenge becomes understanding the interconnected decisions both humans and agents make across the development cycle—if multiple agents triage, research, implement, review, and verify a change, engineers need to reconstruct why that change happened months later, with that record connected to the change itself. The infrastructure play reflects a broader bet that developers will need standardized building blocks to make agentic automation practical at scale, though the path from proprietary backends to industry-wide adoption remains uncertain.