Capabilities
Quick reference for what Loopstack can do. Each entry links to the relevant build guide.
AI & LLM
| Capability | Description | Guide |
|---|---|---|
| Text generation | Call Claude, OpenAI, or other LLMs with prompts or message history | Text Generation |
| Structured output | Force LLM to return data matching a JSON Schema | Structured Output |
| Tool calling | LLM decides which tools to invoke (function calling) | Tool Calling |
| Multi-turn chat | Conversation loops with message accumulation | Chat Flows |
| AI agents | Autonomous tool-calling loops with error recovery | Agent Workflows |
| Multi-provider | Use Claude and OpenAI in the same app, switch per-call | LLM Providers |
| Streaming | Real-time token streaming to the UI | LLM Providers |
| Prompt templates | Handlebars templates for dynamic prompts | Templates |
Workflow Patterns
| Capability | Description | Guide |
|---|---|---|
| State machines | Typed state, automatic persistence, transitions between places | Workflows |
| Human-in-the-loop | Pause for user input, forms, confirmations | Human-in-the-Loop |
| Sub-workflows | Launch child workflows, receive results via callbacks | Sub-Workflows |
| Dynamic routing | Guard-based conditional branching with priorities | Dynamic Routing |
| Error handling | Auto-retry with backoff, custom error places, manual retry | Error Handling |
| State management | Typed state interface, persistence across pauses | State Management |
Tools & Documents
| Capability | Description | Guide |
|---|---|---|
| Custom tools | Reusable logic with Zod schemas, injectable into workflows | Tools |
| Custom documents | Typed UI components with forms, code editors, buttons | Documents |
| Task tools | Tools that launch sub-workflows (for agent delegation) | Sub-Workflows |
Integrations
| Capability | Description | Guide |
|---|---|---|
| OAuth 2.0 | Provider-agnostic OAuth with Google and GitHub built-in | OAuth |
| Docker sandboxes | Run untrusted code in isolated containers | Sandbox |
| Secrets management | Request and store API keys/tokens from users at runtime | Secrets |
| Programmatic execution | Trigger workflows from APIs, webhooks, cron jobs | Programmatic Execution |
UI & Studio
| Capability | Description |
|---|---|
| Visual workflow monitoring | Watch workflow execution in real-time |
| Chat interface | prompt-input widget for conversational UIs |
| Form widgets | Text, textarea, select, slider, code-view, radio, checkbox |
| Document actions | Buttons that trigger workflow transitions |
| Conditional widgets | enabledWhen shows/hides UI based on workflow state |
| Sub-workflow embedding | LinkDocument displays sub-workflow progress inline |
Extensibility
| Capability | Description | Guide |
|---|---|---|
| Custom LLM providers | Implement LlmProviderInterface for any LLM API | Creating LLM Providers |
| Custom OAuth providers | Implement OAuthProviderInterface for any OAuth service | Creating OAuth Providers |
| Registry packages | Browse and install pre-built tools and workflows | Registry |
Tech Stack
- Runtime: Node.js, TypeScript, NestJS 11
- Database: PostgreSQL (TypeORM), Redis (ioredis, BullMQ)
- Frontend: React 19, Vite, Tailwind CSS, Radix UI
- LLM SDKs: Anthropic SDK, OpenAI SDK
- Package manager: npm with workspaces
Last updated on