​
Agent explained

A Restack agent functions as an event-driven background process that persists state, executes tasks over time, and operates continuously for weeks or even months. Restack agents can:
  • Listen for events to run tasks
  • Maintain long-term state to learn and improve
  • Run at scale with built-in scheduling and controls
Built on Kubernetes, Restack agents keep running even when problems occur. They handle heavy workloads reliably. They provide OpenAI compatible endpoints that clients can use to stream responses.

​
Blueprints

This section shows common agent patterns used in production - from basic agents to advanced features like tool calling and RAG.

​
Agent with chat + stream

Restack agents store context for weeks or months without extra setup. They provide OpenAI-compatible endpoints for streaming responses.

Agent with chat

​
Agent with RAG

RAG helps agents give better answers by using proprietary data.

Agent with RAG

​
Agent with tool calling

Tool calling helps control what actions agents can take, making them more accurate.

Agent with RAG

​
Agent with third-party API

This blueprint shows an agent that gets data from external services to create messages with an LLM.

Agent with RAG

​
Agent with PDF extraction

This pattern shows how to handle files like PDF and image. The agent reads the file, gets text through OCR, and organizes it with an LLM.

Agent with PDF extraction