Thanks for the feedback. A few clarifications for anyone reading this thread:
The terms “FREE” and “free tier” refer to Google’s Gemini free quota, not to unlimited usage and not to AI usage being included with Loco AI.
Loco AI is a BYOK (Bring Your Own Key) plugin. We do not provide Gemini access, resell Gemini usage, or bundle Gemini credits into any Loco AI plan.
Our documentation explicitly states:
• “You bring your own AI key. The plugin doesn’t resell AI capacity.”
• “Your AI provider is a separate bill, paid directly to them, never to us.”
• “Google’s free tier is the most generous of the cloud providers.”
• “Once a billing account is attached to your Google Cloud project, requests past the free limits are billed automatically.”
• “Loco AI is BYOK: it sends requests straight to Google and does not cap, meter, or warn on your Google spend.”
Relevant documentation in our website at https://ai.locodev.dev
Regarding API keys: the statement that keys are stored locally and never sent to Loco AI servers refers specifically to API key storage and routing. Requests go directly from the plugin to the AI provider you selected. It is not a statement that AI usage is free.
Regarding billable usage: saving an API key does not itself generate charges. Charges come from requests sent to the AI provider using that key.
Regarding quota warnings, token tracking, and spend tracking: these are not currently features of Loco AI. Provider-side usage, quotas, and billing are managed by Google, Anthropic, OpenAI, DeepSeek, Groq, OpenRouter, etc. We are transparent about this in the documentation and do not claim to provide spend monitoring or quota enforcement.
We take reports of unexpected API costs seriously, but there are a few important distinctions here.
First, the quoted Google support message does not establish that Loco AI caused the charges. Google can see requests made using a Gemini API key and can report usage associated with that key. However, that is different from proving which application, workflow, script, or tool generated those requests.
Second, Loco AI does not have access to your Google billing account, Google invoices, Google quotas, or Google’s internal cost calculations. We cannot independently verify the quoted support response without the original support ticket, request logs, timestamps, and request IDs.
Third, “failed” and “billable” are not mutually exclusive. AI providers bill for API requests that are processed, regardless of whether the user considers the final result useful. This is true across Google, Anthropic, OpenAI, DeepSeek, and other providers.
Regarding retry behavior specifically, Loco AI documents that it has a bounded retry strategy:
“Hard cap of 2 retries on the same approach. After that it stops and asks you instead of spiraling.”
and
“Loco AI has a hard 2-retry rule. If the same approach fails twice, it stops and asks you instead of spiraling into 10 failed calls.”
These behaviors are described in the documentation under “How Loco AI Works” and “Your First Prompt.”
If there is a reproducible case where Loco AI exceeded its documented retry limits or generated unexpected request volume, we’d be happy to investigate it. The most useful information would be:
• Provider request logs
• Timestamps
• Model names used
• Prompt history
• Screenshots or exports showing the reported loop
Without that information, it’s not possible to determine whether the charges were caused by a plugin bug, a provider-side issue, normal API usage, a workflow repeatedly retried by the user, or another source of requests using the same API key.
One thing worth noting is the timeline.
The Google support message quoted in this post references activity from May 14, 2026.
Loco AI has gone through multiple releases since then, including fixes, retry logic changes, stability improvements, and additional safeguards. A report based on behavior from a version that is over a month old does not necessarily reflect the current state of the product.
If a user encounters a reproducible issue on the latest version, we’re happy to investigate it. But using behavior from an older build as evidence that the current version behaves the same way is not an accurate representation of the software today.
If a reproducible bug exists, we’ll fix it. That’s how the product improves.
Technical failures inside Loco AI
The claims about crashes, freezes, failed tasks, and “never successfully completing a single task” are anecdotal experiences, not evidence of a platform-wide issue.
Loco AI is used daily by Unreal Engine developers to generate Blueprints, create gameplay systems, edit assets, and automate workflows. Like any AI-assisted development tool, failures can occur depending on project complexity, engine version, corrupted assets, AI provider responses, or local environment issues.
Importantly, isolated failures do not demonstrate that Loco AI is fundamentally broken or incapable of completing tasks.
The documentation also explains several safeguards specifically designed to reduce common AI failure modes:
-
Deterministic gameplay templates for common systems (Pickup, Damage, Raycast, Enemy, Spawner, HUD)
-
Engine-source verification to validate functions and properties before use
-
Pre-checks that block invalid Blueprint operations
-
Compile and audit loops after edits
-
Retry limits designed to prevent endless failure spirals
These features exist specifically because AI systems can make mistakes, and Loco AI attempts to catch and correct them before they reach the user.
Documentation:
“Every failure triggered more billable API calls”
This statement is misleading.
API providers charge for requests processed, not for successful outcomes.
If an AI request fails to produce the desired Blueprint, the underlying provider (Gemini, OpenAI, Claude, etc.) may still charge for tokens processed because computation already occurred.
This behavior is standard across the AI industry and is not unique to Loco AI.
Loco AI does not control the billing policies of external AI providers.
Billing and cancellation system
The claim that users are “trapped” in subscriptions is not supported by the product documentation.
The official documentation explicitly states:
“You can cancel any time from Dashboard → Billing.”
It further explains:
“Cancellation stops the next renewal but doesn’t refund the current period. After cancellation you stay on Premium until the end of the current billing period, then drop to Free automatically.”
This documented cancellation flow is the opposite of a subscription trap.
If a billing portal was temporarily unavailable or a support issue occurred, that would be a technical problem that should be fixed, not evidence that cancellation is impossible.
Documentation:
Bottom line
There is a significant difference between:
- “I experienced bugs and had a poor experience.”
and
- “The product is fundamentally broken, endlessly loops requests, and traps users in subscriptions.”
The first is a legitimate user complaint.
The second requires evidence beyond individual anecdotes.
The public documentation shows multiple safeguards against AI failure loops and explicitly documents a self-service cancellation path through Dashboard → Billing.
**
Technical-failures safeguards**
Response regarding support, documentation, and transparency
Several claims in this section are contradicted by Loco AI’s publicly available documentation.
“No live support” / “No working contact method”
Loco AI does not advertise 24/7 live chat support, so criticizing it for not providing a service it never claims to offer is misleading.
The documentation explicitly references:
-
Community Discord support for Free users
-
Same-day Discord responses for Premium users
-
Direct contact options for larger teams and studios
A user may have had a support experience they were unhappy with, but that is different from claiming there is “no support” or “no contact method.”
“No documentation explaining costs”
This is factually incorrect.
Multiple documentation pages explain:
-
Loco AI does not include AI usage
-
Users bring their own API key
-
AI providers bill users directly
-
Loco AI never adds markup to AI usage
-
AI costs are separate from the Loco AI subscription
Examples from the docs include statements such as:
“You bring your own AI key.”
“Your AI provider is a separate bill, paid directly to them.”
“We never charge extra on top of what your AI provider bills you.”
“You pay your AI directly.”
The pricing model is described repeatedly throughout the Plans and Pricing documentation.
“No warnings about API usage”
Loco AI does not currently function as an API billing dashboard.
The AI provider (Google, OpenAI, Anthropic, OpenRouter, etc.) remains the source of truth for:
-
token consumption
-
request volume
-
spending limits
-
billing alerts
-
account quotas
This is similar to how IDEs, API clients, and development tools generally work: they send requests, while usage tracking and billing remain managed by the provider.
Not having a built-in spending dashboard is a missing feature request, not evidence of deception.
“No transparency about retries or loops”
This is also contradicted by the public documentation.
The “Why pay when free plugins exist?” page explicitly discusses retry behavior and states that Loco AI includes:
“A retry budget.”
and
“Hard 2-attempt cap on the same failing approach.”
The documentation further explains compile checks, validation systems, and failure handling mechanisms.
Whether those systems worked correctly in a particular user’s project is a separate discussion, but the claim that there is “no transparency” about retries is inaccurate because the behavior is described publicly.
“Users are left unaware that they are generating billable usage”
The documentation repeatedly states that users bring their own AI provider and are billed directly by that provider.
Examples include:
“You bring your own AI key.”
“Your AI charges you directly for what you use.”
“We never charge for AI usage.”
“Your AI provider is a separate bill.”
A user may disagree with how prominently those warnings are displayed, but it is not accurate to claim that the existence of provider billing is hidden or undocumented.
Summary
It’s fair to request better usage tracking, spending dashboards, token monitoring, or stronger warnings about external API costs.
However, saying Loco AI has no support, no documentation explaining costs, no transparency about retries, and leaves users unaware they are using billable AI providers is not consistent with the information publicly available in the documentation. The docs repeatedly explain that users bring their own AI provider, pay that provider directly, and describe several of the systems used to handle retries and failed operations.
Document references:
Regarding the conclusion
I completely support developers sharing negative experiences. If a tool causes problems, those experiences should be discussed publicly so others can make informed decisions.
That said, the conclusions in this post go far beyond the evidence presented.
Statements such as:
-
“The platform is not ready for production”
-
“The integration with AI providers is misleading”
-
“The billing model is dangerous”
-
“The cancellation system is non-functional”
-
“The support system is non-existent”
-
“The technical performance is unstable”
are broad claims about the entire product based on a single user’s experience.
A personal experience, positive or negative, is valuable feedback, but it is not the same thing as demonstrating a platform-wide issue.
Several of the specific claims made earlier in this post are also contradicted by publicly available documentation:
-
The docs repeatedly state that users bring their own AI key and are billed directly by their AI provider.
-
The docs explain the Free and Premium plans in detail.
-
The docs document a cancellation path through Dashboard → Billing.
-
The docs describe retry limits, compile checks, validation systems, and other safeguards designed to prevent runaway AI behavior.
-
The docs clearly explain that AI costs are separate from the Loco AI subscription.
If there are reproducible bugs, billing portal issues, crashes, workflow freezes, or support failures, those should absolutely be reported and investigated.
However, there is a meaningful difference between:
“I had a bad experience and encountered several issues.”
and
“The product is unsafe, misleading, non-functional, and not ready for production.”
The first is feedback.
The second is a conclusion that requires substantially more evidence than has been presented here.
Developers evaluating Loco AI should review the documentation, test the Free plan themselves, verify claims independently, and make their own conclusions rather than relying solely on either marketing material or a single negative review.
Regarding the reports submitted to Google
The fact that a report was submitted to Google Cloud Trust & Safety, Partner Compliance, or an API review team does not, by itself, validate any of the claims made in this post.
Anyone can submit a report to a provider. The important question is whether the provider investigated the report and determined that a policy violation actually occurred.
As written, this section provides no evidence that:
-
Google found wrongdoing by Loco AI
-
Google determined Loco AI violated any policy
-
Google classified the plugin as abusive
-
Google concluded that charges were caused by a defect in Loco AI
-
Google took any enforcement action
Submitting a report is a user’s right and can be appropriate when they believe something went wrong.
However, a submitted complaint should not be presented as proof that the complaint is correct.
If Google, Anthropic, OpenAI, or another provider were to publish findings, issue enforcement actions, suspend access, identify policy violations, or conclude that Loco AI was responsible for improper billing behavior, that would be relevant evidence.
Until then, this section only establishes one thing:
A user was dissatisfied enough to file reports with Google.
It does not establish that the reports were validated, accepted, or resulted in any finding against Loco AI.
Transparency is important. So is distinguishing between allegations, reports, and verified conclusions.
Final response
I appreciate anyone sharing their experience, positive or negative. Developers should absolutely discuss issues publicly so others can make informed decisions.
That said, the central warning in this post:
“You may end up paying for a UI while unknowingly generating billable API usage across multiple providers.”
does not accurately describe how Loco AI is documented.
The documentation repeatedly states that:
-
Users bring their own AI provider.
-
Users bring their own API key.
-
AI providers bill users directly.
-
Loco AI does not include AI usage.
-
Loco AI’s subscription covers plugin features, not AI tokens.
This information appears throughout the pricing, plans, and billing documentation.
A user may disagree with the UX, may want stronger warnings, token tracking, spending dashboards, budget limits, or additional safeguards. Those are valid feature requests.
However, there is a significant difference between:
“I would like more visibility into API costs.”
and
“Users are unknowingly generating billable API usage.”
The latter implies the billing relationship is hidden, when the documentation explicitly explains it.
Ultimately, every developer should do the same thing they would do with any development tool:
• Read the documentation.
• Review the pricing page.
• Understand who bills for AI usage.
• Test the Free plan.
• Verify claims independently.
The Unreal community benefits most when experiences are shared alongside verifiable facts. Potential bugs, crashes, support issues, billing concerns, and feature requests are worth discussing. But conclusions should be based on evidence rather than assumptions about intent or speculation about how the product operates.
Here are all seven contact emails currently published on the site, what each is for, and where it appears:
| Email |
Purpose |
Where it’s shown |
contact@locodev.dev |
General + billing/cancellation (the one I added, one-business-day reply) |
/contact |
privacy@locodev.dev |
Data access, deletion, GDPR/LGPD requests |
/docs/privacy, /privacy, /privacy/telemetry |
support@locodev.dev |
Support |
/privacy/telemetry |
security@locodev.dev |
Security / vulnerability reports |
/privacy/telemetry |
legal@locodev.dev |
Legal |
/terms |
hello@locodev.dev |
General |
/docs/troubleshooting |
Developers should evaluate the product based on current documentation, current releases, reproducible evidence, and their own testing rather than relying solely on any single review, including this one.
