Agent Blueprints
Foundations Workflows Patterns Composition Playground Explorer Compare
Foundations Workflows Patterns Composition Playground Explorer Compare
Foundations Taxonomy

Workflows

  • Evaluator-Optimizer
  • Orchestrator-Worker
  • Parallel Calls
  • Prompt Chaining

Agent Patterns

  • Agentic RAG
  • Event-Driven
  • Long-Horizon
  • Multi-Agent
  • Plan & Execute
  • RAG
  • ReAct
  • Reflection
  • Routing
  • Saga

Primitives

  • Memory
  • Skills
  • Sub-agents
  • Tool Use

Modifiers

  • Guardrails
  • Human in the Loop

Composition

  • Overview
  • Combination Matrix
  • Anti-Compositions
  • Reference Architectures
  • Blueprints → Deployments
  • Blueprint → Spec → Scaffold

More

  • Evolution Explorer
  • Compare Patterns
  • Roadmap
  • Changelog
Workflows

LLM Workflows

Workflows are orchestrated patterns where the code controls the flow. The developer defines the structure; the LLM fills in the content.

When to use workflows instead of agents:

  • The task can be broken into known, predictable steps
  • You need deterministic, auditable behavior
  • Low latency is critical — no dynamic decision overhead
  • You're not yet sure if agent autonomy is needed
Workflow Intermediate

Evaluator-Optimizer

Generate-evaluate feedback loop that iteratively improves output.

Workflow Intermediate

Orchestrator-Worker

LLM decomposes a task and delegates to specialized workers.

Workflow Beginner

Parallel Calls

Concurrent LLM calls on independent inputs, aggregated at the end.

Workflow Beginner

Prompt Chaining

Sequential LLM calls with validation gates between steps.

When workflow logic becomes too complex, it's time to graduate to agent patterns.

Explore the evolution from Workflows → Agents →
Agent Blueprints · MIT License
GitHub Foundations Workflows Patterns Roadmap Changelog Built with Astro