Design Your Agent

Choose your problem first, then get the best pattern recommendation in two quick steps.

Step 1

What do you need help with?

Choose the task area first. Then we will show only matching patterns.

Browse All Patterns

Pattern	Use when	Group	Complexity	Read
Think & Execute5
ReAct Agent	You need iterative tool use where each step depends on new observations.	Think & Execute	Low	Read ->
Task Decomposition Agent	You need to split a broad objective into smaller executable tasks.	Think & Execute	Medium	Read ->
Code Execution Agent	You need the agent to run code, validate outputs, and iterate safely.	Think & Execute	Medium	Read ->
Data Analysis Agent	You need the agent to explore datasets and produce analysis-backed answers.	Think & Execute	Medium	Read ->
Research Agent	You need multi-source research with structured evidence and summaries.	Think & Execute	Medium	Read ->

Multi-Agent & Control4
Routing Agent	You need to route requests to specialized prompts, tools, or sub-agents.	Multi-Agent & Control	Low	Read ->
Orchestrator Agent	You need one controller to plan and coordinate multiple dependent agents.	Multi-Agent & Control	High	Read ->
Supervisor Agent	You need a governance layer that monitors and corrects other agents.	Multi-Agent & Control	High	Read ->
Multi-Agent Collaboration	You need multiple agents to collaborate on different parts of one workflow.	Multi-Agent & Control	High	Read ->

Memory & Knowledge2
RAG Agent	You need grounded answers using external documents at runtime.	Memory & Knowledge	Low	Read ->
Memory-Augmented Agent	You need cross-step or long-term memory to keep task context.	Memory & Knowledge	Medium	Read ->

Quality & Safety4
Reflection Agent	You need a self-review pass before returning the final output.	Quality & Safety	Low	Read ->
Self-Critique Agent	You need explicit critique criteria to catch weak reasoning.	Quality & Safety	Low	Read ->
Fallback-Recovery Agent	You need graceful recovery paths when a step fails.	Quality & Safety	Medium	Read ->
Guarded-Policy Agent	You need strict policy checks before tools or risky actions run.	Quality & Safety	Medium	Read ->