Build my stack
Funding & Deals
Convex Blog: Convex raises a $57M Series B
Aug 4, 2026
- Convex closed $57M Series B led by Insight Partners, with participation from Etna Labs, a16z, Spark Capital, and Justin Kan. Total funding now $110.5M.
- Millions of Convex instances are running in production, with Enterprise plans already adopted by dozens of companies.
- Core thesis: AI agents need a backend that enforces correctness at the platform level, since agents can't review every line they write like human engineers do.
- Funding will accelerate core product development, agentic development tooling including MCP integrations, and hiring at SF headquarters.
Convex, the backend platform founded by former Dropbox infrastructure engineers, announced a $57 million Series B on August 4, 2026. Insight Partners led the round, with Etna Labs, Andreessen Horowitz, Spark Capital, and Justin Kan participating. Total funding is now $110.5 million.
The round comes four years after a16z led the $26M Series A that put Martin Casado on the board. The pitch has evolved significantly: what started as a reactive state-management backend for frontend developers is now positioned as the safe substrate for agent-written software.
Why AI Agents Need a Different Backend
Convex's central argument is that traditional backends trust the developer to get things right. Schema validation, race condition prevention, and data consistency are the developer's responsibility. But when AI coding agents like Cursor, Copilot, or Claude write and deploy backend logic, you need the platform itself to enforce correctness structurally.
Convex provides ACID transactions, automatic reactivity (queries re-run when data changes), and deterministic function execution. If an AI agent writes a mutation that would violate data integrity, the runtime rejects it. No silent corruption, no eventual consistency surprises. The platform catches mistakes the agent didn't.
Traction and Market Position
The company reports millions of instances in production and says revenue has grown substantially since launching Enterprise plans. Their sweet spot is real-time collaborative apps: the reactive data layer means queries automatically update when underlying data changes, eliminating manual cache invalidation and WebSocket plumbing.
Convex competes with Firebase, Supabase, and PlanetScale for developer mindshare. The differentiation is architectural: Convex is not a wrapper around Postgres. It's a purpose-built reactive database where the programming model prevents entire categories of bugs that plague traditional REST + database stacks.
The Bigger Pattern in Developer Infrastructure
Convex isn't raising in isolation. DataBahn closed $40M the same week for an agentic data control plane. Emergent raised $130M at $1.5B for agent-safety infrastructure. HappyRobot pulled in $150M at $1.2B for enterprise AI agents. Investors are clearly paying for the layer that makes agent-written software safe to ship.
The funding will go toward core product development, expanded agentic tooling (MCP integrations, AI-specific SDKs), and scaling the San Francisco team. Expect deeper IDE integrations so coding agents can scaffold full Convex backends without human hand-holding.
Frequently Asked Questions
What makes Convex different from Supabase?
Convex is a purpose-built reactive backend where queries auto-update when data changes. Supabase is a managed Postgres with real-time subscriptions added on top. Fundamentally different architectures with different trade-offs.
Is Convex open source?
Client libraries are open source. The server runtime is proprietary and runs as a managed cloud service.
What's the pricing model?
Free tier for small projects, then usage-based pricing on function calls, storage, and bandwidth. Enterprise plans add dedicated resources, SLAs, and priority support.
Who are the founders?
Former Dropbox infrastructure engineers who built the company's distributed sync engine and storage systems.
Does it work with AI coding tools?
Yes. Convex has MCP server support so agents like Cursor and Claude Code can query schemas, run mutations, and scaffold projects directly.
