What are Agents?

Agents are the core building blocks of Reminix. They're AI-powered services that process requests and generate responses.

The Big Picture

┌─────────────┐     ┌─────────────┐     ┌─────────────┐
│   Your App  │ ──▶ │  Reminix    │ ──▶ │   Agent     │
│             │     │    API      │     │  (Your Code)│
└─────────────┘     └─────────────┘     └─────────────┘

You build agents using our runtime, deploy them to Reminix, and call them via our API.

What Can Agents Do?

Agents handle two types of requests:

Type	Description	Example
Invoke	Process input → return output	"Summarize this article"
Chat	Handle conversation → return response	"Hello! How can I help?"

Agent Components

Every agent has:

Name — Unique identifier (e.g., support-bot, data-processor)
Handlers — Functions that process requests
Server — HTTP server managed by the runtime

How Agents Work

Build — Create an agent with handlers using our runtime
Test — Run locally and test with curl
Deploy — Push to Reminix via GitHub
Call — Invoke via API from your application

Two Ways to Build Agents

Approach	When to Use	Learn More
From Scratch	Custom logic, full control	Python or TypeScript
Wrap Existing	Already using OpenAI, LangChain, etc.	Python Adapters or TypeScript Adapters

Adapters let you wrap agents from popular AI frameworks (OpenAI, Anthropic, LangChain, LangGraph, Vercel AI, Mastra) and serve them with the Reminix runtime.

Example

A simple agent that greets users:

Input: { "name": "Alice" }
        ↓
    [ Agent ]
        ↓
Output: "Hello, Alice!"

Agent Names

Names must be:

Lowercase letters, numbers, hyphens, underscores
Unique within your project
URL-safe (used in endpoint paths)

Examples: my-agent, support_bot, processor123

Next Steps

One-off vs Conversational — Choose the right pattern
Invoke, Chat & Streaming — Request/response details
Python Quickstart — Build with Python
TypeScript Quickstart — Build with TypeScript

On this page