agentgraph CLI is the primary way to execute, inspect, and manage dagraph workflows from your terminal. Every command operates on plain YAML files and local run directories — no hosted platform required. Install the engine package and you have the full CLI immediately.
serve and schedule commands, install the optional extras:
Commands
| Command | What it does |
|---|---|
run | Execute a DAG from start to finish, or resume a previous run |
validate | Parse a DAG YAML, check graph structure, and show the execution wave plan |
inspect | View node statuses and artifact contents for a completed or partial run |
approve | Approve or reject a paused approval_gate node |
respond | Provide a value to a paused user_input node |
list | Show all past runs with status, DAG name, and node summary |
resume | Resume a paused or partial run from where it left off |
graph | Render a DAG as Mermaid or Graphviz output |
diff | Compare node artifacts between two runs of the same DAG |
serve | Start a long-running webhook server that triggers runs on POST requests |
schedule | Run a DAG on a repeating cron schedule |
Global conventions
All commands accept a--runs-dir flag to point at a custom runs directory (default: runs/ relative to your working directory). Run IDs are 12-character hex strings generated automatically; pass --run-id to run or resume to use a specific value.
run
Execute a DAG end-to-end with full backend and sandbox control.
validate
Catch YAML errors and undefined template variables before you run.
inspect
Browse node statuses, artifact digests, and full output content.
approve & respond
Unblock paused HITL gates and user-input nodes.
list & resume
Audit all past runs and resume interrupted ones.
graph & diff
Visualize DAG structure and compare run outputs.
serve & schedule
Trigger workflows via webhook or cron schedule.