Vercel review

Vercel is a hosted cloud platform for building and deploying apps and AI agents. It sits between a codebase and the infrastructure that serves the running application: teams deploy to Vercel, which provides deployment environments, global delivery and infrastructure features including Fluid Compute. Its AI offering adds sandboxed environments, durable orchestration and access to models through AI Model Gateway.

This is a managed platform, not a library or a self-hosted server. The homepage also positions Vercel as a deployment target for coding agents, with an API, CLI, MCP and Skills. That puts it in the delivery and runtime layer of a stack, while an application still needs its own code, data sources and model or agent logic. Vercel says apps and agents can run in sandboxed VMs with durable backends, and highlights hosting for platforms serving customers. It is broader than static frontend hosting, but teams should map each required backend service and its usage charge before moving production workloads.

Vercel names four integration surfaces for coding agents: its API, CLI, MCP and Skills. The homepage says agents can deploy in their native language, but does not name particular programming languages or SDKs. Before choosing an SDK or building a language-specific workflow, confirm current support and examples in Vercel's docs.

The CLI is a direct route into deployment workflows, while the API, MCP and Skills provide other ways for agents to interact with Vercel. The homepage does not describe API conventions, authentication details, webhooks or specific framework integrations, so check the docs for the endpoints and lifecycle events your pipeline needs. The product also offers deployment environments, preview URLs, custom domains and serverless functions as platform features. That can reduce the work of wiring deployment and delivery infrastructure, but teams should verify how their repository, build process and runtime fit Vercel's supported workflow. Documentation depth and SDK coverage matter if you need more than the named integration surfaces.

Hobby is $0 per month for personal, non-commercial projects. Its listed allowances include 1M CDN requests, 100 GB data transfer, up to 3 custom firewall rules, up to 10 IP blocks and 1M ISR reads per month. Pro costs $20 per month for professional developers and teams, includes a $20 monthly usage credit, and lists Flat Rate CDN, up to 40 custom firewall rules, unlimited viewer seats and a free custom domain. Enterprise pricing is custom and includes a 99.99% SLA.

The base plan is only part of the cost when usage-priced features apply. The pricing page lists Fluid Active CPU starting at $0.128 per hour, function invocations starting at $0.60 per 1M, and basic build minutes at $0.007 per minute. AI Gateway token requests cost $3 per 1,000; Vercel Agent is $0.25 per 1M tokens plus pass-through token cost. Storage, image processing, observability and security features also have separate charges. Prices are monthly; annual rates are not listed.

Vercel is presented as a hosted cloud platform, not as a self-hosted server. The vendor doesn't publish a licence or self-hosting path on its product page. Teams considering it should confirm whether any components they depend on are open source and what licence applies before building around them.

The product page also doesn't establish a data-export process. Ask Vercel how to export application data, deployment configuration, logs and any platform-specific state you need to retain. Migration effort will depend on how much of the application uses Vercel-specific deployment environments, serverless functions, Fluid Compute, AI Gateway or other platform services. Some Reddit commenters describe moving away as costly, though those reports are individual experiences rather than a measure of what a migration will cost your team. Keep application logic and data portable where practical, and document dependencies before committing production traffic.

Vercel's global delivery and deployment environments target apps that need managed deployment and hosting. The Enterprise plan lists a 99.99% SLA; teams on other plans should check what service commitments apply to their workload. The pricing page lists security-related options including SAML SSO at $300 per month and a HIPAA BAA at $350 per month. It also lists paid firewall and rate-limit request charges, so confirm how rules and request billing map to your traffic.

The usage model deserves attention in production: compute, function invocations, builds, transfer, storage and observability can all affect the bill. Vercel's pricing page lists individual rates, but teams should model their own traffic and build cadence. Reddit reports range from developers who say Pro works well for small SaaS products to reports of large bills tied to build time or usage. Those are anecdotes, not a forecast. Set billing alerts and review usage before scaling. Ask Vercel about rate limits, incident handling and the exact SLA for your plan.