MonoKit is an AI-driven monorepo toolkit for rapid application development, officially positioned as "Harness as Code." It treats the repository itself as an operating model, helping teams and AI agents build web and mobile products within an inspectable, verifiable system.
Purpose and Core Philosophy MonoKit's goal is to make an application's structure, boundaries, and verification processes clearly readable, so that agents can understand the boundaries and safe operations before modifying code. The official emphasis is to "start from the schema," carrying the schema through validation, OpenAPI, and typed clients, then layering on persistence, testing, and observability, with the harness recording boundaries and checks to ensure workflows are repeatable.
Key Capabilities According to the official documentation, MonoKit includes the following capabilities:
- Typed APIs and workflows: Zod contracts connect validation, interactive API documentation, and typed clients, with built-in request IDs, structured errors, and health checks.
- Persistent workflows: Support for transactional state changes, idempotency, audit entries, and an outbox; signed webhooks and lease-based task processing handle retries and recovery.
- Practical AI engineering: Typed tools, permission checks, structured outputs, and versioned prompts, with deterministic test doubles, evaluations, and tracing.
- One mobile app, two backends: Expo Router connects to tRPC or Hono, sharing session handling, account-scoped queries, and caching rules.
- AI + API unit costs: Bring your own rate card, track usage by request and feature, and explicitly handle unpriced usage to model contribution margin.
- Secrets stored in system storage: dotenvx keeps local keys in the operating system credential store, with three shared environment keys, output redaction, and typed validation covering every app.
- Interconnected authentication: Better Auth provides sessions for Web and Expo, and the Hono API verifies the same identity and enforces ownership on example resources.
- Consistent application surface: Next.js, Expo, and browser extension starter templates using HeroUI, Tailwind, and theme-aware components.
- Content, email, and storage: Sanity content, React Email templates, and AWS SES/S3 integration live in shared packages and can be adapted across apps.
- Polar integration: Provides checkout and license verification patterns as a starting point for paid products,Powered by the user's own Polar account and configuration.
- Deploy on demand: you can start from Vercel, and the Hono backend also includes Cloudflare Workers and Node/Docker adapters, with explicit database migrations.
- Runnable verification: an inspectable harness declares real checks such as type, lint, test, build, and smoke tests, with database fixtures covering concurrency and recovery.
How to use Officially provided structure includes four major application starter templates (apps): api (Hono, supporting Vercel, Workers, and Docker), web (Next.js, tRPC and Better Auth), mobile (Expo, optional tRPC or Hono), extension (Vite browser extension). Shared packages (packages) cover api, ai, auth, database, cms, email, storage, ui, testing, tools, tsconfig, and mcp (a scaffolding tool that defaults to dry-run). In addition, monokit.harness.ts defines application boundaries and executable checks, AGENTS.md provides versioned instructions and guardrails, Envfile defines encryption and output redaction policies, and scripts, docs, recipes, and skills provide shared environment and verification tools, engineering and verification guidance, optional integration references, and reusable agent setup guidance.
Applicable scenarios Suitable for teams that need to build web and mobile products simultaneously and want to base their work on typed APIs, persistent workflows, and practical AI tools, as well as development scenarios that want to provide clear context and executable verification for AI agents.
Limitations and notes The official materials provided do not specify exact pricing, charging details beyond the refund policy, performance metrics, or a complete list of platform compatibility; regarding "What other costs should I expect?" and "Do you offer refunds?" they only mention that a Q&A exists, but do not give specific content. This information is not currently provided and should be confirmed through official channels.