Free automation tools and where an agent still pays

Searches for free automation tools come from people who want to connect steps without paying for a platform. A form submission creates a row, a row triggers a message, a message updates a record. That glue work is where free tools are strongest.

What they rarely include is the data. A workflow that finds a company's size, or checks a search result, or looks up a work email, calls an outside service that charges for the call. This post separates what is free in an automation from what is not, and shows how to add live data to a free workflow tool such as n8n without surprises on the bill.

What do free automation tools usually give you?

Most free automation tools give you one of three things. A hosted free plan with a limit on runs or steps. A self-hosted version you run on your own server. Or an open source core with a paid cloud on top. The details change often, so check each vendor's current terms and note the date.

What you get is the workflow engine: triggers, branching, scheduling inside the tool, and connectors to common apps. For AI automation, many of these tools add a step that calls a language model.

The word free here means the engine is free. You still pay for a server if you host it yourself, and you pay for every outside service the workflow calls.

Where does the cost hide in a free workflow?

In the calls out of the workflow. Think of a lead workflow with five steps: a form arrives, the workflow finds the company, finds a work email, checks the email is valid, and writes a row. The engine is free. Three of the steps call a data provider.

If each provider needs its own account, you now have three signups, three keys, three bills and three sets of terms. The workflow was free to build and is not free to run, and nothing in the engine shows you the total.

How do I add live data to n8n or another workflow tool?

Use an HTTP request step to call a data API, or connect an AI agent step to a tool server. The details depend on the tool, so read its current docs. Two patterns cover most cases.

HTTP step calling a REST API. The workflow sends a request with a bearer token and reads the JSON that comes back. You control the exact call, which suits fixed workflows.

Agent step with tools. An AI step decides which tool to call. This suits open questions, where the steps vary by input.

looot supports both from the data side. It exposes a REST API and an MCP endpoint, and one token covers every provider in the catalog. You do not collect a key for each provider. The price of each endpoint is visible first, so you can set a ceiling in the workflow and stop when it is reached. What 1,000 lookups cost shows how to read the price for a batch.

Should an agent replace the workflow?

Sometimes. A fixed workflow is cheap, fast and predictable, so keep one for a job that never changes. An agent suits work where the steps depend on what it finds: research, list building, checking a result and trying a different source when the first returns nothing.

Many setups use both. The workflow handles the trigger and the writing, and an agent step handles the part that needs judgment. The agent's data calls go through one connection, so the cost shows up in one place.

For examples of agent-run jobs, the enrich a company list recipe and the data enrichment use case show the steps end to end.

What is a free way to test a data step?

Run one real call before you build the loop. Take one company from your list and run the enrichment once. Look at the fields it returns and the price it charged. Then multiply by the length of your list, and decide whether the workflow is affordable before you build the other four steps.

Some free tools on the looot site run real lookups at no charge to you, with rate limits. See free tools for what is there. They suit a single question, not a list.

What should I check in a free plan's terms?

Look for the limit that will stop you first, and it is not always the obvious one. Some plans cap the number of runs. Others cap the number of steps inside a run, the number of active workflows or the time a run may take. A workflow with a loop over three hundred leads can hit a step limit long before a run limit.

Also check whether the free plan allows the connectors you need, and whether a self-hosted version requires you to keep it patched and backed up. Free in money is not free in time. A server you must watch costs attention, and you should count that in the real price of the automation.

What does a five-step lead workflow cost to run?

Take the lead workflow from earlier. The form step and the sheet step cost nothing in data calls. The company lookup, the email lookup and the email check each cost the listed price of their endpoint. Per lead, that is three data calls.

Multiply the number of leads by three, and then by the prices in the table above. For three hundred leads that is nine hundred calls. The cheapest provider for each step gives the low estimate, and a second provider on misses gives the high one. Compare the total with the hours the workflow replaces, and you can decide whether to build it in a free tool or leave it manual. Automation ROI calculator shows the full formula.

Try it: a prompt for your agent

This prompt prices a workflow step before you wire it in. Replace the numbers with your own list size.

The answer gives you a cost before you build anything. No call is run.

What it cannot do

• It does not replace the workflow engine. looot provides data and tool calls, not triggers or branching.
• It does not make a data call free. Each run costs its listed price, and there is no trial credit.
• It does not run on a timer. A workflow that must start by itself needs scheduled runs inside your workflow tool, since looot's own scheduled runs are not available yet.
• It does not check the terms of your workflow tool. Read each vendor's current plan page.

Last checked 2026-10-03 against the live catalog and the looot skill file. Author: Luisa.