One of the first questions people ask about any automation platform is “what will this actually cost me?” And with most tools, the honest answer is: it’s hard to know.

I wanted GloriaMundo’s answer to be different.

The problem with automation pricing

Automation platforms have two common pricing models, and both have problems.

Per-task pricing (Zapier, Make). You pay for each step that executes. Sounds simple until you do the maths. A 9-step workflow processing 200 records per day is 1,800 tasks daily. That’s 54,000 tasks per month. Zapier’s pricing tiers are based on task volume, so a workflow that seems cheap when you’re testing it with 5 records becomes expensive when you’re running it in production. The per-task model also penalises well-designed workflows — adding a validation step or an error handler means paying more, which creates a perverse incentive to build simpler (and less reliable) automations.

Token-based AI pricing (autonomous agents). You pay based on the tokens your agent consumes, but you have no idea how many tokens a run will use until it’s finished. Will the agent make 3 LLM calls or 30? Will it use a cheap model or an expensive one? You find out when the bill arrives. Some agents don’t even show you the cost per run — it’s buried in a monthly aggregate.

Both models share the same fundamental issue: you don’t know what something costs until after you’ve paid for it.

What we do instead

GloriaMundo’s pricing has three parts, and they’re all visible.

$20/month subscription, which includes $20 in credits. This is the base. You can think of the subscription as $0/month with $20 in prepaid credits, or as a $20/month plan where you get your money’s worth in credits. Either way, you start with $20 to spend on actual workflow execution.

2x markup on underlying costs. When your workflow runs, it consumes resources: LLM calls through OpenRouter, integration actions through Composio, image generation, code execution, web searches. Each of these has a real cost that we pay to the provider. You pay twice that. Not three times. Not ten times. Not a variable amount that changes based on your plan tier. Twice.

We show you both numbers. If an LLM call costs us $0.003, you pay $0.006. If a Composio action costs us $0.002, you pay $0.004. The underlying cost and the markup are both visible in your usage breakdown. You can verify that the maths checks out.

Why 2x? Because it’s honest. We need margin to run the platform — servers, infrastructure, engineering, support. A 2x markup covers that while remaining straightforward. Compare this to Zapier, where the relationship between what a task actually costs them and what you pay is… unclear. The per-task pricing is a business model, not a cost reflection.

You see the estimate before you run

This is where pricing connects to the Glass Box principle.

When you run a Virtual Run preview, the cost estimate is part of the output. The CostManager tracks every resource the workflow would consume: LLM tokens (using OpenRouter’s actual reported cost, not an estimate), integration actions, image generation, code sandbox time, web searches. It adds them up and shows you the total.

A typical workflow might break down like this:

LLM analysis step: $0.003 underlying, $0.006 your cost
LLM content generation: $0.002 underlying, $0.004 your cost
2x Composio actions: $0.004 underlying, $0.008 your cost
Web search: $0.001 underlying, $0.002 your cost
Total: $0.010 underlying, $0.020 your cost

Two pence per run. For a workflow that fetches data, analyses it with AI, generates content, and posts it to Slack. If you schedule that daily, it’s about 60p per month in credits.

Even complex workflows with multiple AI steps, image generation, and several integrations typically cost under $0.50 per run. You see the estimate, you decide if it’s worth it, you approve.

Spending controls that grow with trust

New accounts start at Tier 0 with a $20 credit limit. This is a hard cap — if you hit $20 in usage, workflows pause until you top up. It protects new users from accidentally running up a large bill while they’re learning the platform.

As you build payment history, the limits increase:

Tier 0: $20 limit (new users)
Tier 1: $50 limit (after 3 on-time payments of $20+)
Tier 2: $100 limit (after 3 on-time payments of $50+)
Tier 3: $250 limit (after 3 on-time payments of $100+)
Tier 4: $1,000 limit (after 3 on-time payments of $250+)

From Tier 1 onward, you get a 3-day grace period if you hit your limit — services don’t pause immediately, giving you time to top up. Late payment after the grace period drops you down one tier. This is the same trust-building model that credit card companies use, applied to automation spending.

The point isn’t to restrict you. It’s to make sure nobody — including us — can run up a bill that surprises you.

Your credit balance is always visible

In the builder, your current credit balance is displayed at the top of the screen. It’s not hidden in a settings page. It’s not only visible if you go looking for it. It’s there, all the time, so you always know where you stand.

When a Virtual Run estimates that a workflow will cost $0.45 per run and you have $12.30 in credits, the maths is simple and obvious. If a scheduled workflow would exceed your remaining credits, the system flags this during the preview — not after it’s tried to run and failed.

What this means in practice

I’ll be concrete. Here’s what a month might look like for a typical user.

You set up three workflows:

A daily content monitoring workflow that searches the web, summarises findings, and posts to Slack. Costs about $0.03 per run, runs every weekday. Monthly: ~$0.66.
A weekly lead processing workflow that enriches incoming leads, scores them with AI, and routes them. Costs about $0.12 per run, runs weekly. Monthly: ~$0.48.
An on-demand research workflow you trigger manually a few times a week. Costs about $0.25 per run, used 8 times. Monthly: ~$2.00.

Total monthly credit usage: about $3.14. Out of your included $20.

For most users, the included credits cover normal usage comfortably. You’d need to run complex, multi-step workflows very frequently to exceed the included amount. And because you can see the cost before each run, there are no surprises.

The philosophy

Billing transparency is part of Glass Box. If you can see what your agent will do before it does it, you should also be able to see what it will cost before you pay it. These aren’t separate features — they’re the same principle applied to different questions.

I’ve seen too many people get burned by opaque pricing in automation tools. Monthly bills that doubled because a workflow triggered more than expected. Token costs that spiralled because an AI agent got into a retry loop. The fix isn’t cheaper pricing — it’s visible pricing. When you can see the cost, you can control it.

That’s what we built.

Questions about pricing? Want to understand how a specific workflow would be billed? Reach out at gloriamundo.com.