@robota-sdk/dag-cli

Command-line tool for running, validating, and managing Robota DAG workflows — no server required.

Installation

# npx (no install)
npx @robota-sdk/dag-cli run workflow.dag.json

# Global install
npm install -g @robota-sdk/dag-cli
robota-dag run workflow.dag.json

Commands

run — Execute a DAG locally

robota-dag run <file.dag.json> [--env .dag/.env] [--timeout 120000]

Runs a DAG definition in-process. Uses .dag/.env by default for API keys.

validate — Validate a DAG definition

robota-dag validate <file.dag.json>

Checks for unknown node types, dependency cycles, and missing input/output nodes. Exits 0 (valid) or 1 (errors found).

node — Inspect node registry

robota-dag node list                    # list all available node types
robota-dag node info <nodeType>         # full manifest for one node type
robota-dag node schema <nodeType>       # JSON schema for node config

init — Scaffold a new DAG project

robota-dag init

Creates .dag/workflows/hello-world.dag.json, .dag/.env.example, and README-DAG.md.

mcp — Start an MCP server

robota-dag mcp [--transport stdio]

Starts a local MCP server exposing DAG tools to AI agents: dag_nodes_list, dag_nodes_info, dag_run_definition, dag_run_file, dag_validate.

Server mode — Connect to a remote orchestrator

robota-dag run --server --server-url http://localhost:3012 <file.dag.json>
robota-dag definitions list
robota-dag runs status <dagRunId>

Server URL resolution order: --server-url → ROBOTA_DAG_SERVER_URL → http://localhost:3012.

Use with GitHub Actions

Run DAG workflows in CI/CD with npx — no custom Action needed:

- name: Run AI code review DAG
  env:
    ANTHROPIC_API_KEY: ${{ secrets.ANTHROPIC_API_KEY }}
  run: |
    npx @robota-sdk/dag-cli run .dag/workflows/code-review.dag.json \
      --input code="$(git diff origin/main...HEAD)" \
      --output json > result.json

See docs/github-actions.md for full examples (PR review, release notes, scheduling).

Documentation

See docs/SPEC.md for the full package contract.