Choose Your Path
Aerostack offers several ways to build AI-powered backends. Here’s how to decide which one fits your use case.
Quick Decision Tree
“I want to…”
- …expose an API endpoint → Functions
- …build a chatbot on Discord/Telegram/WhatsApp/Slack → Bots
- …give an LLM access to external services → Workspaces + MCP Servers
- …deploy a prompt as a REST API with tool-use → Agent Endpoints
- …orchestrate multi-step logic with conditions and approvals → Workflows (configured inside Bots or Agent Endpoints)
- …let an AI agent manage my infrastructure → Agent Automation
Comparison Table
| Feature | Functions | Bots | Agent Endpoints | Workflows |
|---|---|---|---|---|
| What it is | Edge-deployed REST APIs | AI chatbots on messaging platforms | Prompts deployed as REST APIs | Visual logic orchestration |
| Best for | Custom backend logic, CRUD, webhooks | Customer support, community management | AI-powered API routes with tool-use | Multi-step processes with approvals |
| LLM required? | No | Yes | Yes | Optional (LLM nodes available) |
| Connects to MCP tools? | Via code | Via workspace | Via workspace | Via MCP tool nodes |
| Deploys as | REST endpoint | Bot on 5 platforms | REST endpoint with streaming | Runs inside bot or agent endpoint |
| Setup time | 3 min | 5 min | 5 min | 10 min |
Common Combinations
Customer Support System
Bots (user-facing chat) + Workflows (escalation logic) + Workspace (Stripe, Zendesk, CRM tools)
AI-Powered API
Agent Endpoints (REST API) + Workspace (GitHub, Notion, Slack tools)
Internal Automation
Workflows (scheduled triggers) + Functions (custom logic) + Workspace (external service tools)
Developer Platform
Functions (backend APIs) + MCP Servers (publish tools for other developers)
Still Unsure?
Start with a workspace — it’s the foundation. Connect a few MCP servers, then decide whether to expose them via a bot, an agent endpoint, or a custom function.
Or use the AI Assistant — describe what you want to build and it will set everything up for you.