How the Factory Works

SaaS Factory is live and actively being used: 29 products are running on the platform today, and 257 features have been shipped to this product alone through the autonomous pipeline. That said, we are transparent that the platform is under continuous improvement. Some capabilities — including social media publishing, interactive live sandbox, and certain mobile UX refinements — are built out in the data layer but not yet fully surfaced in the UI. We ship improvements on a regular cadence, and the product you evaluate today will be measurably more capable within weeks.

The platform runs a full team of specialist AI agents — covering research, architecture, implementation, testing, release, marketing, security, CRM health scoring, churn prediction, revenue analytics, support ticket routing, and more. Each pipeline cycle, agents discover feature opportunities, write and test code, open pull requests on your GitHub repository, and deploy via Vercel. You retain control through configurable approval gates: you can require a human sign-off before any merge, or allow fully autonomous shipping. The agents amplify your team's capacity — they do not replace human judgment where you choose to keep it.

Every product runs in its own project scope, and access control is enforced at multiple layers. Role-based permissions govern which team members can read or modify each project. Environment variables and credentials are encrypted at rest. We are actively rolling out row-level security across tenant-keyed database tables as a defence-in-depth measure. GitHub tokens, SSO credentials, and integration secrets are stored with version-prefixed encryption so keys can be rotated without data loss. Full details on our security architecture are available on request.

Every deployment is tracked with a full release record linked to the originating pipeline run and GitHub commit. You can configure approval gates to require human review before any merge reaches production. If a problem does make it through, the release history gives you the commit reference needed to revert — and a one-click rollback UI that triggers a GitHub revert and Vercel redeploy is on the active roadmap. Recurring pipeline failures are automatically surfaced as high-priority features in the queue after three occurrences, so the platform escalates problems rather than silently ignoring them.

The dashboard streams live agent activity via a real-time pipeline detail view, including per-agent job status, log output, and token usage per run. A notification bell in the dashboard header surfaces failed pipelines and pending approvals so nothing goes unnoticed. Token usage and estimated cost data are tracked at the agent-job level in the database. A dedicated cost-breakdown dashboard that shows spend by agent type and by product is on the active roadmap — details available on request for teams with specific budget-visibility requirements.

GitHub is the core version-control integration: agents open branches, commit code, create pull requests, and respond to CI results against your repositories. Vercel handles automated deployment, with environment variable sync and deploy-hook triggering built in. Notification channels include email, Slack webhooks, and Discord webhooks, configurable per product and per event type. Connections to additional services — including OAuth token management and environment variable validation against external APIs — are managed through the platform's encrypted environment settings. An MCP (Model Context Protocol) integration layer to connect agents to a broader ecosystem of external tools is on the roadmap.