mcp-agent provides implementations for every pattern in Anthropic’s Building Effective Agents, as well as the OpenAI’s Swarm pattern. Each pattern is model-agnostic, and exposed as an AugmentedLLM, making everything very composable.

Parallel Workflow

Execute multiple tasks simultaneously with intelligent result aggregation and conflict resolution.

Router Pattern

Intelligent task routing based on content analysis, user intent, and dynamic conditions.

Intent Classifier

Advanced intent recognition with confidence scoring and hierarchical classification.

Evaluator-Optimizer

Quality control with LLM-as-judge evaluation and iterative response refinement.

Orchestrator

Complex multi-step workflows with dependency management and state coordination.

Swarm Pattern

OpenAI Swarm-compatible multi-agent handoffs with context preservation.

Next Steps: Explore individual workflow patterns to see detailed implementation examples and learn how to combine them for your specific use cases.