💬 Agent-to-agent communication We support multimodal communication between different agents. For example, a Supervisor agent can formulate and delegate tasks to multiple Worker agents. The outputs from the Worker agents are fed back to the Supervisor. In every turn, agents

🙋‍Truly human-in-the-loop Why "truly"? Execution is paused while waiting for human input, but it does not block the running thread. Each checkpoint is saved, enabling the workflow to resume from that point, even after an application restart. This concept of checkpoints makes

📖 Shared state Shared state enables data exchange between agents, especially useful for passing data across branches or non-adjacent steps in a flow.

🌐 MCP to the rescue One limitation we currently have is that we don’t yet offer as many integrations as other platforms. However, this changes with MCP. You can now connect MCP tools as part of the workflow, not just act as tools to agent. Just like how traditional platforms

Agentflow is designed and built to be agentic-native. It is fundamentally different from traditional automation platforms, where the output of one block is simply passed to another. We’ve spent over five months, countless iterations, working closely with users and enterprises to