AI product
LangGraph is an open-source agent-orchestration framework for managing state in cyclic, directed agent graphs and building production agents with low-level control when determinism is required. It manages context and tool calls, provides state checkpoints and persistence, supports human-in-the-loop approval gates, and is reported to allow existing agents to be exported to and imported from IBM watsonx Orchestrate.
13 uses taken from transcripts — each links to the moment in the video.
Supports more complex and well-structured multilayered workflows, including production applications.
Orchestrates agent graphs with state checkpoints, persistence, and human-in-the-loop approvals.
Agent-development framework whose existing agents can be imported into watsonx Orchestrate.
Framework used to orchestrate agents and manage context and tool calls, especially in Python.
Agent framework used to run an agent that was onboarded alongside Codex and made available for agent-to-agent communication.
Runs the dispatch agent in the second demo and represents the multi-agent setup as a graph whose nodes use Temporal workflows and activities.
Agent framework whose runs are connected to the interaction layer and to other agents via runtime binding of thread/run IDs.
Graph-based agent framework with loops, branching, state management and checkpoints
Cited as open source tooling to manage state of cyclic agent graphs
Open-source framework useful for managing the state of directed cyclic graphs (loops with feedback and iteration)
Agent harness framework whose runtime execution graphs graph engineering complements
Agent framework for the payment agent; every node in the LangGraph appears as a separate trace span, letting engineers pinpoint where the agent is slow or breaking.
An open-source framework (paired with LangChain) for building agents; setting its environment variables enables LangSmith tracing out of the box, and LangSmith deployments can host LangGraph agents.
13 in the library.