Start free, scale when usage proves it

Yes. The platform is open source under Apache-2.0 and free to self-host: you bring the servers, the Postgres, and an AI provider key (Claude or OpenAI) if you want the natural-language build tools (the self-healing loop itself needs no model). The plans on this page are Backenly Cloud, where we run the infrastructure and handle backups and upgrades. It is the same codebase either way, and pg_dump moves your data between the two.

One credit is 1,000 tokens of Backenly’s own model usage. Credits are spent when Backenly runs its own model on your behalf: the natural-language MCP tool (backend_chat), including the LLM-powered steps inside that turn such as the architect and function generation. They are a small included line, not the headline: the dashboard assistant answers questions free, the typed MCP tools cost nothing, and autonomy never draws credits.

Almost never; it depends which tool your agent calls. The typed MCP tools (create_table, add_column, set_rls, generate_types, run_query and the rest) compile straight to SQL with no model call, so they are free on every plan: your agent supplies the intelligence and you pay your own provider. The exception is the natural-language tools, backend_chat and generate_function, where Backenly runs its own model on your behalf. Those draw credits, and you are billed the tokens the call actually burned, nothing more. Point your agent at the typed tools and the AI meter stays at zero.

Never. Detection and repair are deterministic: probes find drift, and each finding maps to a typed repair that compiles to SQL, with no model in the repair path, so there is nothing to bill you for. When the loop cannot fix something and escalates it to you, a model may write up the diagnosis you read; that pass is on us, never your credits. Every plan gets the identical loop: checks every minute, repairs everything it safely can, with no per-window cap and no monthly limit, Free included. We do not meter healing: a backend that stops repairing itself once you hit a quota is the exact failure we built this to remove. Plans differ on capacity (projects, users, storage, AI credits), never on whether your backend is allowed to stay healthy.

No. Core backend primitives are included on every plan. Paid tiers increase capacity, support, rollback windows and advanced controls.

Your backend is standard PostgreSQL, and it stays yours. Every plan, Free included, gets a real read-only connection string (psql, TablePlus, any BI tool), an optional read-write one, and one-click pg_dump exports that restore on any Postgres: RDS, Neon, your own server. The platform is open source too, so a full self-hosted Backenly is always an exit path. Schema, data, constraints, indexes: everything leaves with you, anytime, with no exit fee.

Yes. Paid access continues until the end of the billing period, then the account returns to the Free plan.

Nothing is billed unless you choose it. Pro’s quotas are shared by every project on your account, and your spend limit is off by default, so each quota is a hard cap. Set a limit ($50, $100, $250 or your own) and pay it in advance; usage then continues at the rates in the table, never past that limit, with emails at 50%, 80% and 100%. Each month’s usage is drawn from what you paid and the rest carries over, and projects, API requests, autonomy, the typed MCP tools, deploys and rollbacks are never billed at all.