Infrastructure · April 28, 2026 · 8 min read
Who Owns the Database for a User-Created AI Company?
Regentics should own the operating records of the AI company. Users should own or connect the product databases their companies build on.
Key takeaways
- The platform database stores companies, agents, issues, approvals, proof, integrations, budgets, and memory.
- Customer product databases can be external resources connected through integrations or repo workflows.
- Everything must stay company-scoped, auditable, and permissioned.
Separate the operating system from the product being built
Regentics needs its own database because it is the control plane. It stores the companies users create, the agents they hire, the issues they assign, the approvals they grant, the integrations they connect, the proof they upload, and the operating memory the company accumulates.
That is different from the database of a product the AI company builds. If a user asks an engineering team to create a SaaS app, that SaaS product may need its own database, migrations, auth, and deployment. The AI company can manage or connect that resource, but it should not be confused with Regentics's internal control-plane data.
The default should be managed and company-scoped
For launch, the platform should manage the operating data in a production Postgres database with strict company scoping. Users should not need to think about that database to create a company.
When their AI company needs external infrastructure, the system should guide them: connect GitHub, choose a repo, create a branch, set up Neon or Supabase if needed, record credentials in secure secrets, and attach the resulting proof to the roadmap.
The CEO should know what exists
The business plan should reflect the real infrastructure state. If the user has a domain, repo, and database, those become company facts. If the database is missing, it becomes a setup issue with a required owner and proof.
That lets the company operate honestly: the platform manages the AI company, while the user controls the assets and databases the AI company uses to produce real-world output.