Convex

What Convex actually is

Convex is a backend platform for TypeScript teams who would rather not assemble a backend out of parts. It bundles a reactive document database, serverless functions (queries, mutations and actions), file storage, full-text and vector search, crons, workflows and authentication into one system where clients subscribe to query results that update automatically when data changes. The pitch is architectural rather than cosmetic: ACID transactions and end-to-end type safety by default, no hand-written API layer, no glue code between database, server and client state.

The company behind it is credible and unusually well-staffed for this problem. Convex was founded in San Francisco in 2021 by Jamie Turner (CEO), James Cowling (CTO) and Sujay Jayakar — early Dropbox infrastructure engineers who built the storage systems behind hundreds of millions of users. Funding has followed the traction: a $3.5M seed in late 2021, a $26M Series A led by Andreessen Horowitz in April 2022, a $24M growth round in late 2025 co-led by a16z and Spark Capital, and a $57M Series B led by Insight Partners in August 2026, bringing total funding to about $110.5M. The August 2026 announcement put scale numbers on the table: nearly 2 million applications from close to 500,000 developers, over 1.2 million weekly npm downloads, roughly 10,000 paying teams as of April 2026, and named customers including OpenAI, Tripadvisor, Solana, Zapier and Reducto.

The lock-in question — always the first one we ask of a proprietary backend — has a better answer here than it did at launch. Convex open-sourced its core backend (convex-backend, some 200,000 lines of Rust with TypeScript tooling) in March 2024, enabling local development and self-hosting on your own infrastructure for compliance-sensitive cases, while the managed cloud remains the production-grade path. The company is SOC 2 Type II certified and HIPAA compliant, and launched European hosting in 2025.

Pricing: what you really pay

Convex's pricing is usage-based in the cloud sense — resources, not seats, drive the bill — with four tiers. Free is a genuine free plan with hard caps (1 million function calls, 20 GB-hours of action compute, 0.5 GB database storage, 1 GB file storage, 40 deployments, up to 6 developers). Starter, introduced in June 2025, is pure pay-as-you-go on top of those free allotments — register a card, pay only for what you exceed, scale to zero — which fixed the old gap where running out of free resources forced a jump to $25/month. Professional is $25 per developer per month and bundles much larger allotments: 25 million function calls, 250 GB-hours of action compute, 50 GB database storage, 100 GB file storage, plus production features like custom domains, preview deployments, log streaming and 24-hour email support. Business (from a $2,500 monthly minimum) and Enterprise (custom MSA, annual commit discounts, bring-your-own-cloud, unlimited developers) add SAML/SSO, dedicated deployments, service SLAs and escalation support.

The unit rates are published in full, which we always appreciate: function calls at $2 per million beyond included (Starter: $2.20), database storage at $0.20–$0.22 per GB-month, action compute at $0.16 per GB-hour on the Convex runtime ($0.32 on the Node.js runtime), file storage at $0.03 per GB-month, search storage at $0.50 per GB-month, data egress at $0.12 per GB. Dedicated deployments (D1024/D2048 classes, from $2,800/month) flip query/mutation compute onto the meter at $0.45 per GB-hour. EU West and Canada Central run 1.3x the US East prices — an honest regional multiplier rather than hidden repricing. Premier support is $1,000/month or 4% of usage; a 24/7 tier with a 30-minute P1 response is $2,000/month or 7%.

The thing to model before you commit: at real scale the meter adds up. Convex itself notes customers paying more than $10,000/month, and the dedicated-hardware path is where the company steers teams that have outgrown serverless economics. Compare your expected function-call, action-compute and storage volumes against those per-unit rates before assuming the developer velocity is free.

Performance & infrastructure

Convex runs as a managed service, and its performance story is about architecture rather than raw hardware specs — there are no vCPU/RAM plans to benchmark against a VPS, so our usual measured-comparison approach does not apply (hosts we have run our dated 2026 sysbench/fio/iperf3 tests on are listed on /benchmarks; Convex is not among them, and we will not invent numbers for it).

What the architecture gives you is well documented. Query and mutation compute is free on serverless deployments — you are billed for function calls and action compute, not for the database reads behind reactive queries, which is generous and shapes how you should structure workloads. Real-time subscriptions run over WebSocket with automatic query invalidation, transactions are ACID and strongly consistent, and functions execute in V8 isolates (a Convex runtime, with a Node.js runtime for actions that need npm packages). Deployment classes scale from S16 (Free/Starter) through S256 (Professional) to dedicated D1024/D2048 classes, and limits are published per tier — 10,000 tables per deployment, 32 indexes per table, 40 deployments on Free/Starter rising to unlimited on Business/Enterprise.

Data residency is selectable at deployment: US East (N. Virginia) at base pricing, with EU West and Canada Central at 1.3x. For teams with European users or data-residency requirements, that EU option (launched alongside the 2025 European hosting push) is the relevant fact. The Components ecosystem — sandboxed, open-source building blocks for auth, workflows, rate limiting, AI file streaming — extends the platform without forking it, and the AI Gateway routes model calls at OpenRouter-equivalent pricing.

Where it falls short

This is still a specialized platform, and specialization cuts both ways. Convex's document model is not SQL; teams with an existing Postgres estate, reporting tooling or analyst expectations will feel the mismatch, and migrations off Convex — while much safer now that the backend is open source — still mean re-implementing reactive query logic elsewhere. The TypeScript-only story is real end-to-end: it is a superb experience if your team lives in TypeScript and a non-starter if it does not (Python and Rust client support exists for calls, but the function-authoring experience is TypeScript-first).

Cost predictability is the second caveat. Usage-based billing rewards efficient code and punishes chatty clients; a subscription-heavy frontend can generate function-call volume that surprises you. We would like to see first-class budget alerts and harder caps on the paid tiers; the docs' limits page helps, but the meter is the meter.

Third, the support SLA is a paid feature. Free and Starter are community and email support; Professional includes 24-hour email response; the 30-minute P1 response times live in the $1,000–$2,000/month premier tiers. For a platform that positions itself on reliability at scale, that distribution is worth pricing into the decision. And while the open-source backend defuses the existential lock-in worry, production-grade self-hosting is a serious operational commitment — most teams will use it as an escape hatch, not a daily path.

Who should pick this

Our pick for TypeScript teams building real-time, collaborative or AI-heavy applications where developer velocity is the scarce resource: multiplayer tools, dashboards, marketplaces, SaaS products with heavy client-side state, and AI agent backends (the platform now markets directly at agentic workloads, and its own tooling — Chef, Components for workflows — reflects that). If your alternative is assembling Postgres plus an API framework plus a sync layer plus search plus file storage, Convex replaces a stack, and the comparison should be against the total cost of that stack, including the engineers maintaining it.

We would steer elsewhere data teams that need SQL and BI tooling, non-TypeScript shops, and anyone whose workload is a better fit for a plain VPS or a traditional database host — see our /category/database-hosting and /category/serverless-platforms listings for the alternatives. And if your usage is heavy but steady, price the dedicated tiers before falling in love with serverless metering.

Key takeaways

• Backend platform (reactive database + TypeScript functions + sync + search + storage + auth) from San Francisco, founded 2021 by ex-Dropbox infrastructure engineers; $110.5M raised through a $57M Series B in August 2026.
• Genuine free plan and a pay-as-you-go Starter tier; Professional at $25/developer/month; Business from $2,500/month; all unit rates published (e.g. $2 per million function calls, $0.20 per GB-month database storage).
• Core backend open-sourced in March 2024 (convex-backend, Rust) with self-hosting possible — the old "not self-hostable" criticism is out of date, though the managed cloud remains the production path.
• US East, EU West and Canada Central selectable regions (EU/Canada at 1.3x); SOC 2 Type II, HIPAA compliant; service SLAs only on Business/Enterprise tiers.
• Compute model is favorable for read-heavy reactive apps (query/mutation compute free on serverless), but chatty clients and action-heavy workloads can run up the meter.
• Best for TypeScript teams replacing an entire backend stack; wrong fit for SQL-centric, non-TS, or cost-predictability-at-all-costs workloads.

Sources

• https://www.convex.dev/pricing
• https://docs.convex.dev/production/state/limits
• https://www.convex.dev/enterprise-pricing.pdf
• https://news.convex.dev/introducing-the-new-convex-starter-plan-pay-for-only-what-you-need/
• https://prnewswire.com/news-releases/convex-raises-57m-series-b-led-by-insight-partners-to-scale-the-reliable-backend-for-the-ai-era-302841686.html
• https://grokipedia.com/page/Convex
• https://probsee.com/p/Convex
• https://ai-market-watch.com/company/convex
• https://startupwiki.tech/startup/convex