Base44

You will like Base44 if you never want to see a server: authentication, a database, payments and hosting come switched on, and the first version of an internal tool exists before the meeting ends. That is the daily trait, and it is the trait for people who do not write code. You will hate it if you want the project in your editor on day one; this is a builder first and a codebase second.

Pick it for the internal tool your operations team keeps asking for. Pick Lovable if you will read the code and want it in a repository from the start.

The risk is the managed backend. Data, authentication, integrations and hosting are all Base44's, invisible by design, so an app that outgrows the platform has no backend to take with it: the schema, the auth rules and the integrations exist as configuration in someone else's system. That is the single-vendor shape at its purest, and most users choose it on purpose because it is the convenient one.

Export the data on a schedule from day one, since leaving later means rebuilding the half you never saw. What it does right: paid plans add code editing and a GitHub integration, so the front end can at least leave.

No benchmark is published. The documented design: 1. Generation runs in a sandboxed execution environment, docker_sandbox listed as true, so generated code is executed rather than merely returned. 2. Users on Builder and above choose the model, Claude Sonnet 5, Claude Opus 5, GPT-5.6 or Base 1, an in-house model with no published evaluation, so the choice is offered without the information needed to make it. 3. Verification is the user viewing the app.

The consequence is that quality varies by a setting the docs do not help you set. The observation: offering four models says the vendor has not found one that reliably finishes.

Wix acquired Base44 in June 2025 for about $80M, and a builder inside a company that already sells sites and hosting to small businesses has distribution the board envies: the buyer's existing customers are the target market, and the sales channel is already paid for. Pricing power comes from a five-tier credit ladder with a 20% premium for monthly billing, which is a parent that knows how to price a subscription.

The acquirer question is answered; the pivot risk is absorption into a Wix menu. Position: durable, buy monthly until you trust the roadmap, and expect the name to change before the product does.

The demo builds an app from a chat, which the operations team will love and engineering will not use. Procurement: Elite at $160 per month is $9,600 for sixty, each holding 1,200 credits, and Enterprise is custom. The useful detail is SSO-enforced workspaces, where removing a user's access terminates their connections immediately, the offboarding behavior the questionnaire asks about and rarely gets. Nothing runs in CI; this is not a developer tool.

Onboarding is a login and a prompt. Approved with conditions: for the operations team, not engineering, and with data export terms in writing.

The interesting bit is backwards: Base44 is not an MCP client, it is an MCP server. Point Claude or Cursor at app.base44.com/mcp, authenticate over OAuth, and you get 13 tools, read_file, write_file, edit_file, run_command and create_checkpoint among them, scoped to one workspace per connection, so my own agent can build inside their platform without their chat window. Everything else is closed: no key, no local model, no source, nothing to fork.

That server is the only part written for someone like me, and it is a good one. Respect for the server only; the rest is rented by the credit.