An agent orchestration framework is a developer library for coordinating agent behavior across multiple steps. It manages how state, tool observations, and memories evolve as an agent plans, acts, revises results, requests human approval, or collaborates with other agents. section 10.1 section 10.4
The framework governs what each step sees, how state changes, how tool outputs enter subsequent model calls, how agents exchange information, and when execution stops. Some frameworks provide machinery for teams to define the agent loop and its boundaries; others provide a pre-assembled harness with established defaults. section 10.4 section 10.7
Long-running coding agent
A long-running coding agent needs explicit state, approvals, and execution control across steps. By contrast, a single retrieval-augmented generation endpoint does not necessarily need multi-agent orchestration; a simpler pipeline may be more reliable. section 10.2
Choose and operate an orchestration framework around the system’s requirements:
- Define context sources, retrieval needs, tool boundaries, memory, state, evaluation, and operational constraints before selecting a tool. section 10.2
- Inspect the context sent to the model, tool outputs reintroduced into it, and state retained across steps. Watch for context drift across agents and tool outputs. section 10.2
- Plan for traces, evaluations, logs, cost monitoring, policy checks, rollback, and safe secret handling. Keep prompts, retrieval configuration, tool definitions, and evaluation data outside framework-specific formats where possible. section 10.2
Table 10.1 A taxonomy of AI frameworks and platforms for context engineering
| Category | Primary role | Examples |
AI application frameworks |
Developer libraries for building LLM applications with different context sources and model integrations. |
LangChain, LlamaIndex, Haystack, Spring AI, LangChain4j, Pydantic AI, DSPy, AI SDK |
Agent orchestration frameworks |
Build agents and coordinate multi-step execution, planning, tool use, agent state, human-in-the-loop control, and multi-agent workflows. Some supply the machinery for building the agent loop, and others deliver it already assembled. |
LangGraph, CrewAI, Agent Development Kit, Microsoft Agent Framework, Agno, Embabel, Parlant, DeepAgents, Claude Agent SDK |
AI application platforms |
Integrated environments for visual design, deployment, monitoring, team workflows, and operational control. |
Dify, LangFlow, Flowise, Zapier, n8n, OpenClaw, Temporal, Amazon Bedrock AgentCore, Gemini Enterprise Agent Platform |
Compare orchestration frameworks with application frameworks and integrated platforms by role and representative tools. section 10.1
