Getting Started

This is the hands-on tutorial that walks you through installing NodeTool, running your first workflow, and reusing it from both Global Chat and a Mini-App. If you need a high-level overview first, read the Start Here flow; otherwise jump right in.

Step 1 — Install NodeTool

  1. Download the desktop build from nodetool.ai for macOS, Windows, or Linux.
  2. Follow the platform instructions in the Installation Guide for prerequisites (GPU/VRAM, drivers, Apple Silicon tips).
  3. Launch NodeTool after installation. Sign in with Supabase, or use Localhost Mode for offline testing (Settings → Authentication).
  4. Install the recommended starter models so the templates run instantly:
    • Open Models → Model Manager from the top navigation.
    • Install GPT-OSS (agentic LLM) and Flux (image generation). Each download shows disk usage; plan for ~20 GB of free space.
    • You can always switch a node to a cloud provider later by pressing the Models button on that node.

You now have a fresh workspace with working models and access to the template gallery.

Step 2 — Run a beginner workflow end to end

Pick one of these templates to learn how nodes connect and stream results. Both live in the Dashboard Templates panel, so you can open them with a single click.

Option A: Creative Story Ideas (text generation)

  1. Open the template from the dashboard and click Open in Editor. The canvas shows:
    • StringInput (prompt)
    • Agent (LLM with planning)
    • Preview (streams the generated ideas)
  2. Customize the input — double-click the StringInput node and enter a few guiding details such as “write story prompts for sci-fi explorers visiting ocean worlds.”
  3. Run the workflow — press the Run button in the lower-right corner or use ⌘/Ctrl + Enter. The Preview node streams each idea as the agent produces them.
  4. Experiment — duplicate the Preview node (right-click → Duplicate) or swap the agent’s model via the Model field to compare outputs.

Option B: Transcribe Audio (speech-to-text)

  1. Open the “Transcribe Audio” template. You’ll see:
    • AudioInput → Automatic Speech Recognition (transcription) → StringOutput.
  2. Add audio — click the AudioInput node, choose Upload, and select a short .wav or .mp3 file (a voice memo works great). The asset lands in the Assets panel for reuse.
  3. Set ASR to your preferred model — use the node’s Model menu to switch between a local Whisper build or a hosted provider such as OpenAI.
  4. Run the workflow and watch text stream through the StringOutput panel. Use Preview nodes on the edge between ASR and StringOutput if you want to see intermediate chunks.
  5. Tidy the canvas — choose Layout → Reset Layout if you experiment with additional nodes and need the Dockview panels back in place.

Whichever option you chose, you now have a working workflow that you can save and trigger from other parts of the app.

Step 3 — Save and re-run it from Global Chat

  1. Save the Creative Story Ideas workflow with a meaningful name (File → Save or ⌘/Ctrl + S). Saved workflows appear on the dashboard.
  2. Open Global Chat (left nav → Chat). This brings up the full-screen assistant described in the Global Chat guide.
  3. Inside chat, click Workflow Menu in the composer, select your saved workflow.
  4. Prompt it — write a prompt that will trigger the workflow from the agent automatically. For example, you could write “write a story about exploring ocean worlds”.
  5. Run it — press the Send button in the lower-right corner or use ⌘/Ctrl + Enter. The Preview node streams each idea as the agent produces them.

Use this loop whenever you want to mix structured workflows with conversational prompting.

Step 4 — Turn it into a Mini-App

  1. Navigate to a workflow.
  2. Click on the Mini-App button in the top-right corner.
  3. Run the Mini-App to verify it behaves the same as the editor version.

Now you have a single workflow that runs identically in the editor, in Global Chat, and as a Mini-App.

Next steps