An open runtime for AI-generated business apps.
rntme turns project blueprints into running business applications. An agent authors a blueprint: domain model, queries, commands, HTTP bindings, UI, workflows, and vendor modules. rntme validates that blueprint in layers and runs it through a standard runtime, so the second, third, and tenth iteration stay consistent instead of becoming one-off generated code.
Star the repo if you want AI-generated apps that are inspectable, repeatable, and built on open contracts instead of closed platform magic.
- Blueprints, not app spaghetti. The project blueprint is the unit of authoring, review, versioning, and deploy.
- Agents author; humans decide. JSON artifacts, fail-fast validation, and stable error codes make generated work correctable.
- Business workflows are first-class. Cross-service orchestration uses BPMN; the current worker targets provisioned Operaton.
- Integrations stay behind contracts. Vendor modules implement canonical contracts for identity, AI/LLM, CRM, storage, and other capabilities.
- Files use a standard storage path. The first storage vendor is an
S3-compatible module with per-service
storage.jsonroutes, direct browser uploads, and conditional bucket provisioning. - Open by default. Runtime, validators, modules, apps, and demos are Apache 2.0. There is no separately licensed commercial layer.
npm install -g @rntme/cli
mkdir my-app && cd my-app
rntme init my-app
rntme skills install --agent claude-codeThen invoke Skill: using-rntme in your agent and let it design the first
service. The CLI also publishes and deploys project blueprints through the
rntme platform:
rntme login
rntme project publish --dry-run --org my-org --project my-app .See apps/cli/README.md for the full CLI surface.
Agents: read AGENTS.md before touching the codebase.
rntme is released under the Apache License 2.0.