Convex has closed a $57 million Series B funding round led by Insight Partners, with participation from Etna Labs, Andreessen Horowitz, and Spark Capital, bringing the company's total outside funding to $110.5 million. The 2021-founded startup, built by former Dropbox infrastructure engineers, offers a backend platform designed specifically to handle code written by AI agents rather than human developers. Convex's core argument is that as language models begin writing applications at scale, traditional loosely typed databases will silently corrupt data, requiring a shift to typed schemas and transactional guarantees.
The company reports its platform now powers close to 2 million applications, with OpenAI and Zapier among its named customers. Convex's product bundles a TypeScript-typed database, ACID transactions, live sync, authentication, file storage, and a retrieval-augmented generation component into a single service, eliminating the need for developers or agents to connect multiple backend tools. Founded in 2021, the company has now raised $110.5 million in total from outside investors. The new capital will fund hiring and platform development as Convex races to establish its category ahead of competitors like Supabase, Firebase, and Neon, which are adding their own agent-friendly features.
CEO Jamie Turner frames the problem Convex solves in specific terms: "agents reach for the first thing that works, and on most backends the first thing that works is subtly, silently wrong," he says, with failures surfacing weeks later in production environments. The report describes a particular failure mode where language models generate CRUD code against loosely typed storage systems—the code compiles and tests pass, but data slowly degrades over time. According to the report, Convex's strategy is to force backends to reject malformed writes immediately rather than accept them and hope for the best, positioning typed schemas and ACID guarantees as essential safeguards once agents build applications at volume.
The bet Insight Partners is making turns on whether typed, transactional backends become the standard foundation for agent-built software, similar to how Vercel emerged as the default platform for framework-first frontends, the report notes. If that shift happens, platforms that help agents produce fewer silent bugs will win, while loose-schema systems that enable Turner's described failure mode will lose ground. Convex positions itself alongside infrastructure like Cloudflare's Kitesurf, rebuilt on the assumption that the caller is code rather than a human. The report flags several gaps in available information: no disclosed valuation or revenue figures, and no breakdown of the 2 million applications between throwaway prototypes and production workloads, which would clarify actual commercial traction. The report also notes that Convex hasn't quantified how much platform traffic comes from agent-generated code versus human-written code using the same SDK, and no independent source has verified the company's figures. For companies deploying AI agents to build or maintain software, the implication is clear: backend choices may need to shift from optimizing for developer speed to optimizing for agent reliability, with strict typing and transactional integrity becoming the price of entry rather than optional features. The infrastructure layer is being rearchitected around a different kind of author, one that doesn't notice when something compiles but slowly breaks.