Balance agent control with agency
Design agents that reliably handle complex tasks with AgentFlow, an agent runtime and low-level orchestration framework.
Trusted by companies shaping the future of agents
How does AgentFlow help?
Guide, moderate, and control your agent with human-in-the-loop
Prevent agents from veering off course with easy-to-add moderation and quality controls. Add human-in-the-loop checks to steer and approve agent actions.
Add human-in-the-loop
Build expressive, customizable agent workflows
Low-level primitives provide the flexibility needed to create fully customizable agents. Design diverse control flows using one framework.
See different agent architectures
Persist memory for future interactions
Built-in memory stores conversation histories and maintains context over time, enabling rich, personalized interactions across sessions.
Learn about agent memoryI've been more productive in the mornings. Could it be the coffee?
Retrieving memory...
You started drinking coffee two weeks ago. Your productivity has increased 23% since then based on your task completion data.
First-class streaming for better UX design
Bridge user expectations and agent capabilities with native token-by-token streaming, showing agent reasoning and actions in real time.
See how to use streamingFoundation: Introduction to AgentFlow
Learn the basics of AgentFlow in this Academy Course. You'll learn about how to leverage state, memory, human-in-the-loop, and more for your agents.
Developers trust AgentFlow to build reliable agents
Build and ship agents fast with any model provider. Use high-level abstractions or fine-grained control as needed.
“AgentFlow is streets ahead with what they've put forward. It sets the foundation for how we can build and scale AI workloads. The next chapter in building complex production-ready features with LLMs is agentic, and AgentFlow delivers an out-of-the-box solution to iterate quickly, debug immediately, and scale effortlessly.”
“AgentFlow has been instrumental for our AI development. Its robust framework for building stateful, multi-actor applications with LLMs has transformed how we evaluate and optimize the performance of our AI guest-facing solutions. It enables granular control over the agent's thought process.”
AgentFlow FAQs
See what your agent is really doing
AgentSmith, our agent engineering platform, helps developers debug every agent decision, eval changes, and deploy in one click.