AI Nodes

Bring LLM intelligence directly into your workflows with prompt-based processing and autonomous agents.

Overview

AI nodes let you integrate large language models into any workflow. Whether you need to classify incoming data, extract structured information, generate text, or let an autonomous agent solve complex multi-step tasks, AI nodes handle it natively.

Watchflows supports both local models and cloud providers. Run models locally with Ollama or LM Studio for complete privacy and zero API costs, or connect to cloud services like OpenAI, Anthropic, and Google Gemini for access to the most capable models.

All AI nodes support {{variable}} interpolation in prompts and goals, so upstream payload data flows seamlessly into your AI interactions.

Nodes

Provider Setup

Before using AI nodes, configure at least one AI provider in the Watchflows settings. Providers define the endpoint, authentication, and available models for your AI nodes to use.

Watchflows supports Ollama, LM Studio, OpenAI, Anthropic, Google Gemini, and any OpenAI-compatible API endpoint. See the AI Providers concept page for full configuration details.

Key Capabilities

  • Variable interpolation — reference any upstream payload data in prompts using {{variable}} syntax.
  • Structured output — enforce JSON schema validation on AI responses for reliable downstream processing.
  • Local-first — run models entirely on your Mac with Ollama or LM Studio for complete privacy.
  • Provider flexibility — switch between local and cloud models without changing your workflow structure.
  • Autonomous agents — let an AI reason through multi-step tasks with tool use and self-verification.