LangGraph + DataRobot

What you edit in the example: e2e-tests/dragent/langgraph/myagent.py (graph + prompt) and e2e-tests/dragent/langgraph/workflow.yaml (LLM wiring + workflow._type). DRAgent runs the pair.

Installation

pip install "datarobot-genai[langgraph]"

Guides (what appears in the sample)

Doc	Focus
agent.md	`workflow.yaml` + what `myagent.py` defines
hitl.md	Human in the loop: `interrupt`, resume, checkpointer, `thread_id`
mcp.md	Extra tools merged into your graph (when MCP is enabled)

Env reference: LLM configuration (shared).

Run the e2e sample

From e2e-tests/:

export DATAROBOT_API_TOKEN=YOUR_DATAROBOT_API_TOKEN
export DATAROBOT_ENDPOINT=https://app.datarobot.com/api/v2
uv sync --group dragent-langgraph
uv run --group dragent-langgraph nat dragent run \
  --config_file dragent/langgraph/workflow.yaml \
  --input "Say hello in one short sentence."

HTTP: e2e-tests/dragent/Taskfile.yaml.