Learning Lindy

Action Configuration

Everything you need to know about configuring actions.

An action is a specific operation Lindy can perform, such as “Search the Internet”. Actions include first and third-party tools that Lindy uses to execute tasks.

Examples of first-party actions are Lindy Meetings, Lindy Mail, and Lindy Embed. Third-party actions integrate with external applications. Examples include Slack Send Direct Message, Create Calendar Event and Gmail Send Email. Actions added in Lindy’s Flow Editor and follow a trigger.

Most actions will offer the following configuration fields:

Account Selection
Model
Ask for Confirmation‍
Parameters
- AI Prompt
- Set Manually
- Auto

Account Selection‍

Account Selection lets you select the account you want the action to use.

If you authorize your company Gmail account in the account selection step, Lindy will only have access to your company email. Another example would be for a Calendar event trigger, if you authorize your personal Google Calendar account for the trigger, Lindy will not wake up for events on your business Google Calendar.

Lindy will always ask you to authorize the account if you have not already done so. You can authorize multiple accounts for a given application, but each trigger or action will only have access to one account.

Model

This is where you select the model to be used for the action. The model will be used to complete the parameters of the given action, using context it has about the workflow and data received from the prior steps of the execution. Bigger models are smarter but more expensive.

Lindy provides the top AI models ready to use right out of the box, with no need for model API integration or additional setup.

We recommend starting your Lindies with bigger models until you get their high-level flow right, and then to gradually downgrade their models to optimize your costs. If no model is selected, the action will default to the Lindy’s default model found in the Lindy-level Settings.

You can read more about models and their costs here.

Note, if a field in a given action is in “Set Manually” mode, there is no AI model being used to complete the field.

Ask for Confirmation

This field lets you insert a human in the loop at any step in your Lindy’s workflow.

The Ask for Confirmation field only appears for actions with "side effects," e.g. writing / updating data in other apps.

If this is turned ON, Lindy will not take action (send the email, update the record, create the Calendar event, etc) until you approve the action in the Lindy task view.

To “approve” the action before Lindy proceeds, you will see the action as a “draft” in your Lindy’s task view (where you see the history of workflow executions). Once you click “Send email” (or whichever action was set to “Ask for Confirmation”), Lindy will proceed with executing the task.

Example email action card below for when a Lindy has “Ask for Confirmation” turned on:

When your Lindies have tasks that are pending or require confirmation, there will be a notification in your Lindy’s sidebar on the given task.

Parameters

There are three settings for all action fields:‍

1) AI Prompt

When you set a field to "AI Prompt," you can provide natural language instructions (a prompt) to Lindy on how to complete the field.
Lindy will use the prompt you give her in the field and the context of the previous steps, including data from prior triggers and actions to complete the field.

Guide to crafting effective prompts here.

You can reference previous steps in your workflow in the prompt to give Lindy more direction as to what you want her to complete for the field.
You can do this by using natural language and refer to the step like so, “use the email address found on the New Row Added” or “summarize the content you found on the webpage and draft 5 tweets based on the content”

Let’s see how the AI Prompt, works! We’ve set an Email Subject and Body content using AI Prompts referencing previous steps in this workflow:

Lindy used the content from the previous webpage search and the instructions in the Subject and Body prompts to craft an email to John.

2 ) Auto

When you set a field to “Auto,” Lindy will infer what to type in this field:
- When Lindy completes a field, she has visibility over everything that happened in the automation prior to the action you are configuring.
- For example, if she just performed a Web Search in the previous action, she will see the list of results in this step. If you have left a Google Sheets Update Row action on Auto, Lindy will try to update the row values according to the relevant column headers and search results found from the web search.
- Pro tip: If you are having Lindy update a sheet/table/record, ensure you have explicit headers, as Lindy will see those and infer what cells to update, given the header title and context she has from the steps prior.

"Auto" is best used when the fields you want Lindy to complete are highly variable and are only determined at the time of the execution of the task.

For example, if you want Lindy to update a Google Sheet with information she receives from an internet search and email body content, but are unsure of what information will be extracted from either action, you can leave the Google Sheets Update Row fields on "Auto."

This allows Lindy to update the rows using intelligence based on the context available at runtime.

3) Set Manually

When you set a field to “Set Manually,” Lindy will use the raw value that you specify in the field.
- If you set the CC field in the Send Email action to “Set Manually” and type your email address in the field, Lindy will always CC you in this action.

An easy way to think of the difference between “AI Prompt” and “Set Manually”:

If you type “the current date” in a field that is in “AI Prompt” mode, Lindy will enter the current date — for example, Jan 1st 1970.

But if you type “the current date” in a field that is in “Set Manually” mode, Lindy will literally type “the current date” in this field.

Because "Set Manually" does not use an LLM, it consumes less credits. Substantially less.

If you refer to trigger or action output data while in "Set Manually" mode, this will help optimize your credit consumption as data is transferred directly from one step to another.

If you do not require an LLM for a given field, use the "Set Manually" mode. Because there is no AI involved when Lindy executes a field set to "Set Manually", you will incur less costs referencing output data.

If you set a field to "Set Manually" and leave it blank, Lindy will omit this field.

‍Referencing Trigger and Action Output Data

In both the “AI Prompt” and “Set Manually” settings, you can reference prior trigger and action output data.

For example, if you received a Slack message containing information about an alert, you can reference the Slack message text in the Body field of your Send Email action.

When in “AI Prompt” mode, Lindy will use the prompt in combination with the data, to complete the field.

When in “Set Manually” mode, there is no LLM call or AI involved in completing the field. This is particularly useful when you want to reference the exact data from prior steps in your automation.

‍

Example below:

First, Lindy will craft a Slack Message using the AI Prompt in the Message field of the Slack Send Direct Message action.

Slack Direct Message is received:

The following Gmail Send Email actions Body field is in “Set Manually” mode and refers to the Slack Send Direct Message [message.text] from the step prior.

The exact [message.text] body was sent as an email body. Because the Subject field was in “Auto” mode, Lindy inferred the best subject line suited for the email and completed the field for the user.

Guides

Building a Chatbot: 101

Meeting Notes: 101

Lindy Prompt Guide

Calling any API

Meeting Notetaker with Follow-Up Email‍

Marketing Focus Group