Cookbook

How agents use tools and run in loops

Build a retrieval-augmented generation agent

Create sessions, read state, and manage lifecycle from a plain script

Create a loop-mode session and inspect workflow state with getLoopState

Choose which built-in Coder agents are enabled for a session and set the default agent

Save skills, group them into buckets, and attach them to a Coder session

Store reusable Coder selections in a workspace and attach them to sessions

Find live or wakeable sessions and prepare one for remote attach

Subscribe to a session's live event stream, page through its event history, and hydrate a client from the durable replay stream

Recursive research loop with native tool calling

Run long-running tasks asynchronously

Persist conversation history

Schedule tasks that read and write data

Type-safe API calls between server and React frontend

Use LLMs to evaluate agent outputs

Common server-side utilities

Build a semantic product search

Configure Tailwind CSS for your frontend

Headless browser in a sandbox for AI-guided browsing

Handle incoming webhooks

Deploy Mastra agents with tools, structured output, and multi-agent workflows

ReAct agents, dynamic tools, and streaming with LangChain.js

Tool calling, handoffs, and streaming with the OpenAI Agents SDK

Conversational code intelligence with sandbox execution

Multi-platform chatbots for Slack and Discord

Add agents to an existing Next.js application

Add agents to an existing TanStack Start application

Add Agentuity agents as a workspace package in a Turborepo monorepo with shared types