Frequently asked questions
Answers about building new systems, legacy modernization, production controls, pricing, security, and the API.
Bondi is an AI-native platform for building, running, and continuously evolving custom business systems. You describe the operation in your own words; what you get back is a real PostgreSQL system — tables, screens, permissions, workflows, integrations and AI steps — versioned together, with every change reviewed before it reaches Production.
A board or a table view holds a process until the process gets specific. Bondi generates the model underneath: real tables and relations, permissions derived from the screens people actually use, workflows with typed failures, and an API surface generated from your own schema. You describe the data, the rules, the exceptions and the AI boundaries, and you approve the change set before anything runs.
No, and most teams should not. The common shape is Bondi running the operational layer — the intake, the exceptions, the approvals, the portal — next to an ERP or accounting system that stays the book of record. Be clear about what connecting means: fourteen built-in connectors with typed actions and signed events, an HTTP node, and a published SDK for the system nobody ships a connector for. There is no CRM, ERP or accounting connector and no two-way sync engine, so the pattern that works is one system of record per object, connected deliberately.
Two different sizes of thing. A whole operation — client delivery, orders and inventory, customer onboarding, bookings — built as a system with its own model, screens and roles. Or a single job that currently costs somebody an afternoon: reading the purchase orders that arrive as PDFs, sorting a shared mailbox, generating and sending a document, holding an approval that today lives in a thread. The second kind does not require replacing anything you already run.
Not for the building path. What you do need is someone who understands how the business actually operates and can review the proposed model, rules, permissions and exceptions — because the builder asks where your description leaves a decision open, and an unanswered question blocks the plan rather than being guessed at. Legacy migrations additionally need implementation expertise, and are scoped as a project.
Through a controlled modernization path, one bounded flow at a time, with the rules recovered before anything is rebuilt and a way back from each cutover. That is different from building a new system and is scoped and quoted as an implementation project, separately from the subscription. The assessment can also conclude that replacing is the wrong move.
Start a 30-day trial without a card. The trial begins when you create your first Workspace, and a dedicated database exists from that first workspace. Enterprise and legacy modernization begin with a conversation instead.
Each customer gets a dedicated PostgreSQL database rather than a shared table with a tenant column. Traffic is served over HTTPS/TLS, data at rest carries provider platform encryption, credentials live in an encrypted secrets store separate from the workflows that use them, and backups are kept in a dual-region European archive with public access blocked. The security page states the mechanism and the boundaries rather than an adjective.
Judge it on the mechanisms, not on the assurance. Each environment is its own database; Pro adds Development as a branch of Production with reviewed publishing, version history and version restore; workflow runs are written to your own database, one row per node; and sensitive changes keep an audit history. Contractual uptime targets and service credits exist only under an Enterprise agreement — where there is no contract, there is no SLA, and we would rather say that here.