# Lindy AI: How to Build Your First Agent Step by Step *How-to — 2026-10-05 — by Mahmoud Zalt* A practical walkthrough of building your first agent in Lindy: the workflow menu, triggers, actions, testing, and what each task costs in credits. **Short answer.** **You build a Lindy agent by creating a workflow, giving it a trigger, then adding the actions it should run.** In the dashboard, click the chevron next to the New chat button and choose New workflow, then either pick a template or start from scratch. In the flow editor you add a trigger, add actions under it, toggle the workflow on, and watch runs in the Tasks view. **About this guide.** This guide is based on publicly available information and Lindy's own documentation, checked on 17 August 2026. Sistava is not affiliated with Lindy. Interfaces and pricing change often, so confirm the current steps in their documentation before you rely on them. Most people arrive at Lindy with one job in mind. They want an assistant that watches their inbox, or drafts the same reply they type twenty times a week, or files something into a spreadsheet without being asked. The product can do all of that, but the first hour is spent learning a vocabulary: workflows, triggers, steps, actions, agent steps, tasks, credits. This guide walks that first hour so you do not have to guess. Before you open anything, write down the single job in one sentence. Something like "when a customer emails asking about pricing, draft a reply and notify me." Lindy's own best practice guidance leans the same direction: start small, pick one time sink, let it own that, then keep adding. A vague goal produces a vague agent, and you will spend the afternoon debugging your own indecision rather than the software. ## What is a Lindy agent made of? A custom agent in Lindy is made of workflows, steps, integrations, and memory. The workflow is the job description. The steps are the individual moves inside it. Integrations are the connections to your other tools. Memory is what lets the agent keep past context and preferences when continuity matters. Inside a workflow, Lindy's documentation names six kinds of step. Knowing these six is most of the learning curve, because every screen in the flow editor is some combination of them. - **Triggers: **the event that wakes the agent up and starts the workflow. - **Actions: **the specific operations the agent performs, like sending an email or creating a calendar event. - **Agent steps: **the step where the AI decides for itself how to complete something more open ended. - **Integrations: **the connections to outside systems that actions run against. - **Conditions: **branching logic that decides which actions run. - **Looping: **running the same work in parallel across a set of items. ## How do you create a new workflow in Lindy? You create one from the top left of the dashboard. Click the chevron next to the New chat button and select New workflow. A modal opens offering popular templates, including a Meeting Notetaker, a Lead Generator, and an Email Responder, plus a Start from scratch option and an Explore templates button for the full gallery. Take the template on your first build if one matches your job. A template arrives with a working trigger and a plausible set of steps, so you spend your time editing something that already runs rather than assembling from an empty canvas. Start from scratch is the better choice once you know what each step does and the templates keep giving you pieces you delete. 1. **Open the new workflow menu** — In the dashboard, click the chevron next to the New chat button in the top left and choose New workflow. 2. **Pick a starting point** — Choose a template such as Meeting Notetaker, Lead Generator, or Email Responder, or click Start from scratch for a blank canvas. Explore templates opens the wider gallery. 3. **Add the trigger** — In the flow editor, click the plus sign or right click the canvas and add a trigger. This is the event that wakes the agent up. Authorize the connected account if the trigger needs one. 4. **Add the actions underneath** — Click the plus button on the canvas, choose the action from the integration library, then configure its parameters and select which connected account it should use. 5. **Add judgment where you need it** — Use conditions for clear branching rules, and an agent step where the work needs the AI to decide how to proceed rather than follow a fixed path. 6. **Turn it on and watch it** — Toggle the workflow on so it runs whenever the trigger fires, then monitor executions in the Tasks view. ## How do triggers work in Lindy? A trigger is the event that wakes the agent up and starts the workflow. Lindy's documentation groups them into three kinds: time based, which fires on a schedule such as every weekday at 9 AM, chat based, which fires when someone starts a chat, and event based, which fires when something happens in a connected tool, like a Slack message arriving, an email landing, a calendar event, a new spreadsheet row, or a webhook. The part worth learning early is the trigger filter. A filter puts conditional logic at the trigger itself, so the agent only wakes for events you care about instead of waking for everything and then deciding. Filters match without regard to case, combine with AND and OR, and support nested condition groups. Calendar style triggers also expose a minutes offset, where a negative value fires before the event and a positive value fires after it. One agent can carry more than one trigger. Several triggers can feed the same workflow when different events should start the same process, or an agent can hold separate workflows behind separate triggers. You connect an extra one by adding a new trigger and linking it to the steps you want it to enter. ## How do you add and configure an action? You add an action by clicking the plus button on the canvas, choosing it from the integration library, then configuring its parameters and the account it runs against. The two configuration fields you will meet on almost every action are account selection, which picks the connected account, and a model label, which picks the AI model used to fill in the parameters. Lindy's docs separate basic actions from linked actions. A basic action is a single operation like sending an email or creating a calendar event, and it costs at least one credit per task. Linked actions create connected downstream steps with more than one path, and a particular variant called a channel listens for future events such as a reply to the email your agent just sent. An agent can hold several channels at once, which is how one agent keeps separate conversations going in parallel. **The other way to get this done.** Wiring a trigger to an action and then testing the path is exactly the work we removed at Sistava. Rather than assembling steps, you hire an AI Employee and tell it in plain English what you want handled. The tools are connected already, it asks you when it is unsure, and it sets itself up. There is no flow to build and nothing to maintain when the job changes. ## How do you test a Lindy agent before trusting it? Run it against real data before you leave it alone. Independent walkthroughs of the builder describe a Test control that executes the workflow using your actual inbox content, and they recommend trying more than one scenario, including the cases your agent should refuse to handle. If your workflow branches, you have not tested it until every branch has run once. After it is live, the Tasks view is where you see what actually happened. Check it daily for the first week. Most first-build problems are not broken software, they are a trigger with no filter firing on everything, or an action pointed at the wrong connected account. ## What does a Lindy agent cost to run? Lindy meters work in credits, and the unit that consumes them is the task. Their documentation defines a task as everything the agent does after being triggered, including the conditions, actions, agent steps, and messages that follow. Most tasks cost 1 to 3 credits on basic models and around 10 credits on large models, and every task costs at least 1 credit. On the pricing page we read on 17 August 2026, Lindy listed Plus at $30 per user per month with 3,000 credits per user, Pro at $100 per user per month with 15,000 credits, Max at $200 per user per month with 35,000 credits, and Enterprise on custom terms. The same page described plan credits refreshing monthly and not rolling over, and admin top-ups at $10 per 1,000 credits in blocks of 1,000, with purchased credits carrying forward. Prices move, so read their page rather than this paragraph before you budget. **The gotcha the tutorials skip.** Credits run out mid-month, and when they do, agents pause and new tasks are put on hold. An unfiltered trigger is the usual cause: an agent that wakes on every inbound email burns credits on messages it was never meant to touch. Put the filter on the trigger, not in a condition three steps later, so the agent never wakes for work you do not want to pay for. ## Is there a way to use Lindy without building a workflow? Yes. Alongside the builder, Lindy's quickstart documents a setup where an admin installs Lindy from Slack, teammates connect their own account, and you then connect a Google or Microsoft account for email and calendar, add a phone number or Apple Messages, and link tools such as Notion, HubSpot, or Linear. From there you ask in plain language, for example what is on your calendar today, and it answers. Their best practice guidance for this mode is worth borrowing whatever tool you use. Tell it the outcome rather than the steps, give it context about your role, working hours, and timezone on day one, and correct it quickly when it gets something wrong. That is closer to onboarding a new hire than configuring software, and it is the mental model that produces better results from any of these products. It is also the same bet Sistava made, taken further. You hire an AI Employee for a role, brief it in a paragraph, and it works from that brief rather than from a canvas you maintain. If you find yourself enjoying the flow editor, stay with the builder, it is genuinely good at what it does. If you find yourself avoiding it, that is a signal about which shape of product fits you. ## FAQ ### Do you need coding skills to build a Lindy agent? No. The builder is visual. You add a trigger, add actions from an integration library, and configure each step through forms rather than code. What it does require is clear thinking about the job: what event should start it, what should happen next, and what should happen when the situation is not the normal one. ### How long does building a first Lindy agent take? If you start from a template that matches your job, expect well under an hour for a simple single-trigger workflow, most of it spent authorizing accounts and adjusting the trigger filter. Building from a blank canvas with branching logic and several integrations takes considerably longer, and testing every branch is the part people underestimate. ### What is the difference between a workflow and an agent in Lindy? In Lindy's own terms, a custom agent is made up of workflows, steps, integrations, and memory. The workflow is how the job gets done, defined in the flow editor. An agent can hold more than one workflow, each with its own trigger, so one agent can cover several related jobs. ### What is an agent step and when should you use one? An agent step is where the AI decides for itself how to complete something rather than following a fixed sequence. Use one when the input varies too much for conditions to cover, such as interpreting a messy inbound message. Use plain actions and conditions when the rule is knowable, because they are cheaper and more predictable. ### How much does a Lindy task cost in credits? Lindy's documentation states that most tasks cost 1 to 3 credits on basic models and around 10 credits on large models, with a minimum of 1 credit per task. A task means everything that happens after the trigger fires, not each individual step, so a long workflow is still one task. Check their current credits page, since these figures change. ### What happens when you run out of Lindy credits? Their documentation says agents pause and new tasks are put on hold when credits are exhausted, and that plan credits do not roll over between billing cycles. Admins can top up with additional credits. The practical defence is filtering your triggers so agents only wake for work that matters. ## Sources - [Lindy documentation, ](https://docs.lindy.ai/fundamentals/lindy-101/create-agent)Create an Agent - [Lindy documentation, ](https://docs.lindy.ai/fundamentals/lindy-101/introduction)Lindy 101 Introduction - [Lindy documentation, ](https://docs.lindy.ai/fundamentals/lindy-101/triggers)Triggers - [Lindy documentation, ](https://docs.lindy.ai/fundamentals/lindy-101/actions)Actions - [Lindy documentation, ](https://docs.lindy.ai/account-billing/credits)Credits - [Lindy documentation, ](https://docs.lindy.ai/pricing)Pricing, read 17 August 2026 - [Lindy documentation, ](https://docs.lindy.ai/start-here/quickstart)Quickstart - [Lindy documentation, ](https://docs.lindy.ai/start-here/best-practices)Best Practices - [FlowHunt, ](https://www.flowhunt.io/blog/how-to-use-lindy-ai-beginner-guide/)How to Use Lindy AI, a beginner's guide **Tags:** lindy-ai, ai-agent-tutorial, workflow-automation, no-code-automation, ai-agents-for-business