Trigger nodes start the workflow. Every workflow has exactly one trigger, and nothing runs until it fires. The trigger determines when the workflow runs and what data it starts with. Triggers are covered in detail in the Evo Workflow Automation: Triggers article.

Action nodes do the work. Everything after the trigger is an action node - the steps strung along the line that send emails, query data, call services, make decisions, and transform information. Most action nodes follow a simple pattern: take the previous node's output, do one thing, pass the result on.

Sub-nodes configure a parent node rather than running on their own. They attach to an anchor on the side of a parent node - most commonly the AI Agent - and define how that parent behaves. A sub-node does not sit in the line and does not execute independently.

Send email - sends an email to one or more recipients, with support for dynamic content built from earlier nodes in the workflow.

HTTP request - makes a request to any web address. Use this to call an external service, retrieve data from an API, or send data to a system that does not have a dedicated integration node.

Respond to trigger - sends a response back to the caller of a Webhook trigger. Use this when you need full control over what the caller receives, including the message content and format.

Execute workflow - calls another workflow and waits for it to finish before continuing. Use this to reuse a sub-workflow from multiple parent workflows.

Sticky note - adds a note to the canvas for documentation purposes. Sticky notes never execute and are skipped by the engine.

Access LLM — routes through Access's AI platform. No separate account or API key is needed. This is the simplest option for most workflows.

Anthropic — connects directly to Anthropic's Claude models using your own API key.

OpenAI — connects directly to OpenAI's models using your own API key.

Google Gemini — connects directly to Google's Gemini models using your own API key.

Other providers — additional LLM providers are available for teams with specific requirements.

Simple memory - stores conversation history for the current run only. Suitable for single-session conversations.

Chat memory manager - maintains history across multiple runs of the same chat session. Use this when the agent needs to remember what was said in earlier sessions.

Knowledge base search - searches a knowledge base you have set up in Workflow Automation and returns relevant content.

HTTP request - calls a web address. The same node that acts as a step in the line elsewhere becomes a tool the agent can call mid-reasoning.

Evo tool - gives the agent access to capabilities from other Evo products.

Code - runs a calculation or data transformation the agent can use as part of its reasoning.

Web search - searches the web and returns results for the agent to use.

You want to reuse an agent that is already built and maintained elsewhere.

The agent's capabilities are shared across multiple workflows or products.

You want to keep the workflow canvas simple rather than assembling an agent from sub-nodes.

You need to start the workflow - use a Trigger node.

You need to send something, call a service, or write a record - use an Action or Integration node.

You need AI to analyse, generate, or decide - use an AI Agent node.

You need to reshape, filter, or store data - use a Data node.

You need to branch, loop, or control the flow - use a Logic node.

You need to connect to an external service - use an Integration node.

You need to reuse an agent built elsewhere - use the Agent Builder node.