Developer experience for agent-assisted teams
A codebase your team and your agents can navigate
CentraKit treats AI-assisted development as a product requirement: clear boundaries, explicit contracts, quality gates, and documentation that tells agents where to work.
Architecture proof
A full product stack with boundaries agents can respect.
The point is not just to ship a first screen. It is to give agents and engineers the same map across product, API, docs, CLI, and provider code.
SaaS app included
Start from a real multi-tenant SaaS app with auth, organizations, roles, onboarding, and customer-facing workflows.
API plus MCP server
Pair the product app with a dedicated API, committed OpenAPI output, typed client generation, and MCP for agent clients.
CLI and docs
Keep the product surface broader than the browser with a CLI and docs app that explain the stack to customers, teammates, and agents.
Swappable data layer
Use Supabase by default, but keep providers behind clean contracts so you can swap to Convex or your own backend later.
The repo tells the story clearly
CentraKit stays legible for humans and agents by separating contracts, provider implementations, runtime wiring, and app surfaces.
apps/ saas product app api rest api + mcp server docs product and api docs cli centrakit command surface packages/ data zod contracts + service interfaces core runtime wiring supabase provider implementation ui shared shadcn-style components
Modern developer experience
A stack tuned for human review and agent execution.
CentraKit leans into the tooling choices that make full-stack work easier for humans to review and easier for AI agents to modify safely.
Next.js 16 with cache components, PPR, auth interrupts, and streaming-aware page structure.
A Turbo monorepo that ships the product app, API, docs, CLI, and shared packages together.
TypeScript-first package boundaries with shared Zod contracts and typed clients.
Oxlint, Oxfmt, and Knip for a cleaner default workflow than the usual lint stack.
Centralized tests for unit, integration, and end-to-end coverage.
Rules, skills, and agent context shipped with the repo so AI-assisted development scales with the codebase.
A verification path your team will actually use
The template bakes in commands for formatting, linting, typechecking, OpenAPI generation, dead-code detection, and tests so agents can verify the same path as your team.
pnpm dev pnpm dev:showcase pnpm openapi:generate pnpm lint pnpm typecheck pnpm verify
That same structure gives coding agents better context: clear ownership, explicit service layers, and fewer chances to wander across the wrong boundary.