3 Startups · $26M+ · AI Frameworks + Learning + Sales Automation

What they're building

Eighteen months ago, Sam Bhagwat and his co-founders walked out of a New York AI hackathon bothered by the same question: why were developers building AI agents in Python notebooks when the web runs on JavaScript? LangChain, AutoGen, and CrewAI all started in Python. Their JS ports arrived later, as afterthoughts. The result was a fragmented, inconsistent experience for the hundreds of thousands of TypeScript developers who build the products that most people actually use. Mastra is built from scratch for that audience. It is an open-source TypeScript framework that gives developers the full primitive stack for building production AI agents: workflow orchestration for complex multi-step operations, agent memory across sessions, tool integration, RAG for knowledge, streaming responses, evals for quality measurement, traces for observability, and a Mastra Studio interface for interactive debugging. The architecture is designed for progressive disclosure, simple things are simple, complex things are possible, and everything works together rather than requiring developers to glue libraries from different ecosystems. As of the Series A, Mastra is simultaneously launching a platform layer (Mastra Studio, Mastra Server, Memory Gateway) for running agents at scale, extending from framework to infrastructure. Enterprise features (RBAC, SSO, ACL) require a commercial licence. The open-source core handles everything up to that tier.

Why this matters

The seed round story is exceptional and worth understanding as a hiring signal context. Over 120 investors participated, including Paul Graham personally, Guillermo Rauch (who built Vercel, the deployment infrastructure that runs most of the modern JS web), and Amjad Masad (Replit), alongside every major name from the Gatsby and Netlify ecosystems. That is not a cap table that forms around a pitch. It forms around a team's demonstrated ability to build things developers love. Gatsby at its peak was used by hundreds of thousands of developers and ran on essentially every major content website. The people who backed Mastra saw what Gatsby accomplished and bet that the same founder with the same co-founders and the same community instincts would achieve category definition in agent infrastructure. Spark leading the Series A confirms the thesis is tracking: Mastra has moved from a framework developers use to a platform that production engineering teams at scale-ups depend on. Brex, the fintech recently acquired by Capital One for $5.1B, used Mastra to build agents internally. MongoDB, Workday, and Salesforce are using it for DevOps and SRE automation. Replit is embedding it so users can build agents within the product. That breadth, both developer tooling and enterprise platform, is the sign of a framework that is genuinely becoming infrastructure.