Architecture overview

LogicGen is organised around outcomes, not tools. You describe what you want to happen. The system decides what needs to be built, assembles it, and connects the parts — without you specifying each piece individually.

• Input: a plain-English description of a business outcome.
• Output: a surface, a workflow, and a campaign — connected.
• Control: two review gates before anything goes live.
• Record: every action logged automatically.

Everything in LogicGen is built from six core objects. They nest inside each other from workspace down to deployment.

• Workspace — the top-level container. Holds clients, members, and connected tools. One workspace per business or agency.
• Client — an isolated unit inside an agency workspace. Has its own projects and access controls. Not used in single-tenant workspaces.
• Project — a named initiative inside a workspace or client. Groups related Lead Automations together.
• Lead Automation — the core delivery unit. One surface, one workflow, one campaign. Built from a single description.
• Integration — a connected external tool (CRM, email, payment, etc.). Scoped to the workspace and used by Lead Automations inside it.
• Deployment — the live instance of a Lead Automation. Created on approval. Immutable; revisions create new deployments.

Every Lead Automation moves through six stages. The two review stages are the only moments where human input is required.

• intake — you describe the outcome. LogicGen asks clarifying questions if needed.
• plan — LogicGen produces a structured plan: surface type, workflow steps, campaign shape. You review and approve or revise.
• generate — LogicGen builds all three layers from the approved plan. No further input needed.
• review — you inspect the output before it goes live. Request changes to return to generate.
• deploy — on approval, the Lead Automation is deployed. Integrations activate. The surface goes live.
• monitor — activity is recorded. Conversion events, integration actions, and errors appear in the audit trail.

Workspace boundaries are hard. No data, credentials, or customer records cross workspace lines unless explicitly exported.

• Integrations are scoped to the workspace that connected them.
• Client workspaces inside an agency cannot see each other.
• Members can be invited into specific workspaces or clients — not both by default.
• Lead Automation templates can be copied across workspaces, but live data stays isolated.

This model makes LogicGen safe for agency use — a single team member can manage multiple clients without accidental data bleed.

Data moves through a Lead Automation in a predictable direction: inbound from the surface, processed by the workflow, outbound via integrations, and stored in the deployment record.

• Surface captures input — form fill, click, payment, or page visit.
• Workflow processes it — scores, routes, enriches, or stores.
• Integrations execute — CRM write, Slack alert, email trigger, or webhook.
• Deployment record stores the event — queryable from the audit trail.

Lead Automation surfaces are deployed to a global CDN on approval. Workflows run as serverless functions. Campaigns are handed off to the connected email or outbound tool.

• Default deployment target: Vercel. Bring your own domain or use the LogicGen subdomain.
• Data storage for form responses: Supabase or your own database via integration.
• Deployments are immutable. Revisions produce a new deployment. The previous one is preserved in history.
• Rollback is available from the audit trail — re-activate any previous deployment in one action.

Every stage transition, approval, deployment, and integration action is appended to an immutable log. The log is scoped to the workspace and can be filtered by object or event type.

• Recorded events: plan created, plan approved, generation started, review opened, approved, deployed, integration triggered, deployment rolled back.
• Each event records the actor, timestamp, and affected object.
• Logs are exportable as JSON or CSV.
• Read more at Audit Trail.