Run
AgentFlow
The execution runtime for agent graphs
AgentFlow executes graphs of agent nodes, tools, and edges — coordinating multiple agents within a single flow, with fault tolerance and state persistence for work that runs longer than a single prompt. It's the runtime Agent Builder is built on.
Python SDK · commercially licensed · runs entirely on your own hardware
What it does
Fault-tolerant execution
Run state persists, so long-running agent work survives failures instead of starting over.
Built-in working memory
Agent context and intermediate results are held in memory across steps in a flow.
Tool-native orchestration
Native support for MCP tools, so agents call tools without custom glue code.
Observable execution
Every run streams events, so you can see what an agent is doing while it runs — not just the result.
Multi-provider
Works across OpenAI, Anthropic, Qwen, Gemini, DeepSeek, and other LLM providers.
Load or build flows
Load exported flows from Agent Builder, or construct graphs of nodes, tools, and edges programmatically.
Get started
AgentFlow ships as prebuilt wheels on GitHub Releases, not on PyPI.
# download the wheel for your Python version and platform, then
pip install <wheel file>
See the releases page for available wheels, or Agent Builder for the no-code interface built on top of it.
Let's talk about your agents
Whether it's agent memory, a no-code build, or in-house model serving, tell us what you're working on and we'll point you to the right place to start.
Free Architecture Review
We'll look at your current agent setup and flag where memory, tooling, or serving is the bottleneck
Custom Deployment Guidance
Tailored recommendations for your use case, team size, and hardware
Enterprise Support Options
24/7 support, priority updates, and dedicated engineering resources
Built for teams running agents in production
From open-source Lint-AI users to enterprises deploying custom inference — RooAGI covers the whole stack.