Terms of service.

Omnial MCP is a product of Pixel Labs Solutions LLC, a limited liability company formed in Georgia, USA. Your card statement will show Eden Engine — the trading name Pixel Labs Solutions LLC uses for this service — and Pixel Labs Solutions LLC is the company you are contracting with when you use this service.

See contact for how to reach us.

In these terms, “we”, “us” and “Omnial MCP” mean Pixel Labs Solutions LLC. “You” means the person or business using the service, and “your agent” means whatever software calls it on your behalf. By creating an account or making a call, you agree to what follows.

Omnial MCP is a directory of paid, third-party tools that an AI agent can search, inspect, learn how to use, and call over the Model Context Protocol. When your agent calls a tool, we route the call to the third-party provider that operates it, hold the estimated cost against your prepaid balance, and charge you what that call actually cost; see billing below.

Nothing runs unattended on your behalf outside a call your agent explicitly made. There is no human review of a run before it executes, and no human writing the input; that is the product.

You sign up with an email address; there is no password. Signing in works the same way, by a one-time link emailed to you. You are responsible for the inbox that link goes to and for everything done through your account and its API keys.

Age

You must be at least 18 years old to use Omnial MCP. We do not knowingly collect personal information from anyone under 18 and we are not currently able to verify age at signup, so this is an undertaking you are making to us. If you believe a child has created an account, email pixellabsweb3@gmail.com and we will delete it.

API keys

A key is shown to you once, at creation, with the scopes and spend cap you gave it. You are responsible for keeping it secret; a key without the runs:execute scope cannot make a paid call at all, and a capped key can only lower its own cap, never raise it or mint an uncapped successor.

What it takes to make a call

Creating an account and browsing the catalog costs nothing. Before a paid call runs, your workspace needs a payment method on file and enough available balance to cover the quote; the system checks both first and refuses the call, before any provider is contacted, if either is missing.

Your input

You keep everything you send us. You give us permission to store it, process it, and pass it to the provider whose tool your call names, for as long as we need to run the service, bill it, keep the ledger accurate, investigate a suspected breach of these terms, and comply with law; see third-party output for what happens to it once it reaches that provider.

There is no subscription and no seat to buy. You top up a balance with a card through Stripe; the minimum top-up is $5.00. Your agent spends against that balance one call at a time, and every cent that moves is an entry in an append-only ledger you can read from the dashboard.

On top of what a call costs the provider, we add a platform markup (the same policy across the catalog, applied by the pricing engine before your agent is ever quoted a number). The exact figure for a specific call has one authoritative source: calling the tool with dry_run: true, which prices the call without running it. See pricing for the full mechanism.

Auto top-up

If you enable it, we refill your balance by card when it falls below a threshold you set. The refill amount, a cooldown after a successful charge, and a daily ceiling on automatic charges are all yours to configure, and all exist so a declined card cannot retry-storm your own card in a loop. You can disable auto top-up at any time.

Payment

Card payments are processed by Stripe. We never see or store your full card number; Stripe holds it. The statement descriptor shown at checkout and on your card statement is Eden Engine; the merchant of record is Pixel Labs Solutions LLC. Prices are in US dollars and exclude any tax that may apply where you are. If a payment fails or is reversed, we may suspend new calls until it is resolved.

Credits themselves

• Balance does not expire while your account is open.
• Balance has no cash value outside Omnial MCP, cannot be transferred between accounts, and is not a deposit or stored-value instrument.
• We may change the markup or a provider's price for future calls. A price change never retroactively drains balance you already hold.

A provider being listed in the catalog is not an endorsement of everything it returns. Before you rely on a result (especially anything about health, finance, law, safety, or a named person or business), check it. If a provider's behaviour is wrong in a way that looks like our bug rather than that provider's own answer, tell us what happened; we would rather hear about it than have you assume it is unfixable.

Each provider operates under its own terms of service and its own privacy policy, which govern how it handles the input you send it once your call reaches it. We do not control that provider's infrastructure, cannot audit its internal practices, and do not guarantee its availability. Review a provider's own terms and privacy policy (linked from that tool's usage doc where the provider makes them available) before you send it anything sensitive.

Do not use Omnial MCP, or ask a tool through it, to produce or facilitate:

• Sexual content involving minors, or any sexualised depiction of a real person who has not consented.
• Content that harasses, threatens, defames, or incites violence against a person or group.
• Impersonation: presenting a real person, business, or public official as saying or endorsing something they did not.
• Deliberate misinformation about elections, public health, or an ongoing emergency.
• Infringement of someone else's copyright, trademark, or likeness rights.
• Instructions for building weapons or synthesising harmful substances, promotion of self-harm, or fraud.
• Anything unlawful where you are, where the provider operates, or where the output will be used.
• Circumventing a provider's own usage limits or terms, or a scope, cap, or rate limit we place on your key or workspace.

You also may not, with the data a provider's tool returns to you through Omnial MCP:

• Build any product out of it that resembles or competes with any product of that provider — or of any data source behind the tool.
• Give your own users programmatic or bulk access to it — an endpoint, export, feed, or query surface over that data — in any way that would let them reach a substantial set of that provider's data fields — or of any data source behind the tool — or a material portion of either one's underlying data set. Your own product consuming that data internally to complete a task your user asked it for is not that, however often it does so; handing your users the data itself to query is. However much of that internal use happens, the restriction on building a competing product above still applies.
• Use it, resell it, license it, make it available, or authorise or permit anyone else to use it, for any purpose that provider's own acceptable-use or data-use policy — or that of any data source behind the tool — prohibits, or otherwise act contrary to those policies.
• Use it to decide, or to help anyone decide, a person's eligibility for credit, insurance, employment, housing or tenancy, education, a licence, or a government benefit or service; to take any adverse action against a person; for debt collection, a background check, or any purpose governed by the Fair Credit Reporting Act; to re-identify pseudonymous data by joining it to personal information; for the benefit of a single political candidate, party, or campaign; or as training data for a generative model.
• Use it to promote or advertise firearms, ammunition, or weapons; adult content or services; credit-repair services; payday-loan services; or tobacco or marijuana products; to promote the purchase of a particular company's stock; to benefit a government intelligence agency; or for any evidentiary purpose related to law enforcement or criminal prosecution.

If you are not sure whether a specific product of yours falls inside this, ask us. We will tell you how we read the restriction as it applies to your case; that is guidance, not a waiver or a change to the restriction itself.

Omnial MCP is two layers, and this section treats them differently. The server layer is Omnial MCP's programmatic surface: the tools themselves, called with your own API key, from your own product, backend, or your own AI agent — including from inside your own agent harness, a commercial one included — to do your own product's own work. That is meant to be integrated, and is permitted, including offering the resulting feature to your own users under your own brand with no Eden Engine branding visible to them: a commercial agent-harness product calling our MCP server to run a tool as one step in a task it is doing for its own user, with nothing in that product naming Eden Engine or Omnial MCP, is exactly the intended use, needs no special agreement, and needs no conversation with us first — the same self-serve signup and API key every integration uses is the only door. One limit on that: if you are reselling access to these tools as a product in its own right, rather than using them to power your own product's features, talk to us first. That is not covered by the permission above. Reselling access at scale without a Server-Layer Expansion Agreement — the separate signed instrument described below — is a breach of these terms.

The prohibited uses listed above apply to the server layer as well; the limit on reselling access at scale described just above is about when you need to talk to us first, not the only thing you may not do.

The agent layer is different: the catalog itself — the tool listings, the hand-authored usage documentation, the course content we publish on this service, and the browsable, searchable presentation of them — is what we are, not plumbing, and no part of the permission above extends to it; the paragraphs below state exactly where that line falls, and the whitelabel section of /why walks through how it works in practice, with worked examples.

You may not resell, republish, redistribute, rebrand, or create a derivative work from the agent layer, or from the course content we publish on this service — whether or not you have users of your own to offer it to. Making any of it available to your own users as something they browse, search, or read through your product is one way of doing that, and it breaches this the moment it makes the agent layer, in whole or in substantial part, what you offer them, whether or not you compete with us. Using the agent layer is licensed to you; offering it is not, and nothing in these terms licenses that. The license in ownership below is what licenses that use — reading, searching, and acting on the agent layer as part of your own integration — and it stops exactly there: it is not a license to resell, republish, redistribute, rebrand, or make the agent layer available to your own users as something they browse, search, or read through your product. The permission above reaches the server layer only and grants nothing about the agent layer either way; the written-agreement exception stated further below reaches the server layer only too and can never widen what you may do with the agent layer. We do not offer the agent layer for resale, redistribution, or rebranding and do not intend to; if that policy ever changes, it will take a new, signed version of these terms that says so in as many words, not anything reachable through the exception below. That holds however you get there: taking a copy of it, storing or caching it as something your users browse or search instead of coming to us, and proxying it live through your own agent so your users see the catalog under your brand are all “being us with a different name,” not integration — your users seeing every tool and every usage doc that way is one example of that, not a threshold you must clear before it counts. A derivative work includes a paraphrase, translation, or summary of the listings or the usage documentation that you publish, store, or make available to your users as a substitute for coming to us for it (a summary produced in the course of running one task for one user is not that), a machine-generated adaptation of either, and using either as training data for a model you offer to third parties.

Not a clone: a commercial agent-harness product whose agent calls our MCP server to execute a tool as one step in a task it runs for its own user — including calling omnial_search and showing that user the results so they can pick one in reply to a request they made for that task, or naming a tool and what running it will cost, in the course of that one task — where the harness itself does not stand up a browsable, searchable surface over our catalog. That is the server layer, permitted regardless of how large the harness is or whether it is sold commercially. A clone: an agent harness that wires its own search-and-lookup surface over our catalog so its users can go to that harness, instead of to us, to find, compare, or read about tools, even if browsing is a small part of what the harness does. The moment your users can browse our catalog through your product as a destination in its own right, you are offering the agent layer, not calling the server layer, however small that surface is.

Where the line actually falls. What is prohibited is standing up a browsable, searchable surface over our catalog inside your product: an interface, a listing page, a standing picker, or a stored index your users go to in order to find, compare, or read about tools. What is permitted, and never counts as offering the agent layer, is the ordinary running of a task: your agent naming the tool it is about to call, its provider, and what the call will cost; quoting or summarising as much of a usage doc as that one task needs, to that one user, in the course of doing it; and caching listings, schemas, or prices for your own integration's performance. Telling your user what your agent is doing and what it costs is disclosure, not distribution, and we would rather you did it. The test is whether your users can go to your product to browse our catalog instead of ours — not whether our content ever appears on their screen.

A separate written agreement can widen what the server layer permits — a higher call volume, a dedicated environment, or terms tailored to a bulk or enterprise integration — but only if it is signed by an authorized officer on each side, and only if it expressly names this exception: it must say, in terms, that it is a Server-Layer Expansion Agreement under this section of these Terms, and nothing else qualifies. Accepting these Terms, creating an account, making a call, or exchanging emails with us about a possible deal is never that agreement; a negotiation is not a license, and only a signed document that says so is. As stated above, no agreement of any kind reaches the agent layer. If you want to talk about a server-layer arrangement like this, the whitelabel section of /why explains how to reach us.

You also agree not to systematically scrape or bulk-extract any part of the service, including catalog or course content — including in order to reconstruct or republish it elsewhere — and not to reverse-engineer the service. Ordinary use of the public catalog API to look up a tool, and ordinary search-engine crawling of these pages, are the intended public surface of this product and are not what those restrictions mean. You are responsible for what you ask a tool to do. If a refusal seems wrong to you, tell us which call and why.

The catalog listings as we compile, structure and describe them, the hand-authored usage documentation, and the course content we publish on this service are ours: we write them, and we own them.

Using Omnial MCP gives you a limited, revocable license to use that content: to search and read the catalog, to call the tools it describes — including having your agent receive the usage documentation through omnial_learn as part of that call — as part of your own integration, and to read the course content for your own learning. It does not give you a license to resell, republish, redistribute, rebrand, or create a derivative work from the listings, the documentation, or the course content themselves, or to make any of them available to your own users as something they can browse, search, or read through your product; see acceptable use above for exactly where that line falls.

None of this makes the catalog secret. The listings themselves, and the course content, are deliberately public, indexable by search engines and readable by anyone with no account at all (see acceptable use) — the license above is about what you may do with the content, not about who may look at it. The usage documentation is different: we do not publish it on these pages or let search engines index it. It is delivered to your own agent, byte for byte, through omnial_learn once you are calling the tool it describes — a narrower audience than the listings, not a wider one.

These pages are published, not dedicated to the public. The catalog listings as we compile, structure and describe them, and the course content are protected by copyright, whether or not you have an account with us, and reading them is free. The usage documentation is equally ours and equally protected by copyright; it is simply delivered to an authenticated agent rather than published on a page, and reading it that way costs nothing beyond having an account. Copying, republishing, or building a product out of any of it — the listings, the usage documentation, or the course content — is not free, and the restrictions in acceptable use apply to anyone who does so, account or no account, to the fullest extent the law allows.

We may suspend or close an account, an API key, or access to a specific tool that breaches these terms: particularly the acceptable-use rules above, an attempt to circumvent a spend cap or scope, a chargeback filed without contacting us first, or conduct that risks the accounts we hold with our own providers (Stripe foremost among them). Where practical we will tell you why.

If we close your account for a reason that is not your fault (we discontinue the service, for example), we will refund your unused balance. There is no automatic mechanism for this; it is a Stripe refund to your original payment method that we process by hand once we close the account. If we close it for a serious breach, we may apply your unspent balance against amounts you then owe us that are due and liquidated — a negative balance, an unpaid charge, or a chargeback fee — and refund the rest. We will not withhold your balance against a claim for damages we have not yet established. See the refund policy for how a refund actually reaches you either way. A run already charged and delivered stays charged regardless of why the account is later closed.

We may also remove a provider tool from the catalog at any time, including immediately, if a call against it costs more than 2x its quote, which we treat as our defect and not yours.

You can close the account, any time, from the Close account control on your billing dashboard, or by calling POST /v1/billing/close-account with a key that holds the keys:admin scope. Closing revokes every API key on the workspace, detaches your saved card from Stripe, and turns off auto top-up; no key can make another call, and no new top-up can be started. It refuses while a run is still in progress or an auto-top-up charge is in flight; try again once that clears. Closing also tries to expire any Stripe Checkout Session it can already tie to a top-up you started, but a checkout opened moments before closing may not be tied to one yet, and can still complete; if the expiry call to Stripe fails for some other reason, closure proceeds anyway rather than waiting on it. Either way, a payment that completes is credited to your balance regardless; it lands on a closed account exactly like any other balance and is recoverable the same way.

Closing is not deletion. Your workspace, its run history, and its append-only ledger all stay exactly as they are (we do not erase them on closure), and any remaining balance is neither refunded nor removed; it stays recorded exactly as it was. See the refund policy for what that means if you were hoping to get an unspent balance back.

A closed account can be reopened. Closing revokes every API key on the workspace, so no key can call the API to reopen it either; an account can only be reopened from the Reopen account control on your billing dashboard, while signed in. There is no waiting period. Reopening makes your remaining balance spendable again (less anything refunded or reversed while the account was closed), which is what "neither refunded nor removed" is actually kept for. Reopening does not restore anything else: every revoked API key stays revoked (mint new ones after reopening), your saved card stays detached from Stripe, and auto top-up stays off until you configure it again.

To have your account and its data actually deleted, rather than just closed, email pixellabsweb3@gmail.com; see the privacy policy for what deletion removes and what it does not. Deletion, unlike closure, is not something we can undo.

The service is provided “as is”. To the fullest extent the law allows, we disclaim all implied warranties, including merchantability, fitness for a particular purpose, accuracy, and non-infringement.

Specifically, we do not promise that: the service will be uninterrupted or error-free; any specific provider tool will remain in the catalog or available at a given price; a provider's output will be accurate, complete, or fit for your purpose; or that any particular result will be achieved.

Some jurisdictions do not allow certain warranties to be excluded. If you live somewhere that gives you rights that cannot be waived, nothing here removes them.

To the extent the law allows, neither party is liable to the other for indirect, incidental, special, or consequential damages, or for lost profits, lost revenue, lost data, or loss of goodwill, including anything arising from a third-party provider's output being wrong, unavailable, or removed from the catalog, or from how a third-party provider handles, stores, secures, or otherwise processes data you send it. That exclusion does not apply to a breach of the acceptable use or ownership sections above, or to your obligations under the indemnity below. For those, every category of loss listed in this paragraph — including lost profits, lost revenue and loss of goodwill — remains recoverable: cloning the agent layer costs us exactly that kind of damage, and this paragraph does not sign it away.

Our total liability for any claim relating to Omnial MCP is limited to the greater of $100 or the amount you paid us in the 12 months before the claim arose. This limit does not apply to returning unspent balance we owe you; that is your money, not damages — and, as stated in billing, it has no cash value outside Omnial MCP and is not a deposit or stored-value instrument.

Nothing here limits liability for fraud, for death or personal injury caused by negligence, or for anything else that cannot lawfully be limited.

A breach of the acceptable use or ownership sections above — building what we are, rather than integrating what we offer — causes us harm that a later payment cannot undo. You acknowledge and agree that any such breach would cause us immediate and irreparable harm, that money damages alone would not be an adequate remedy for it, and that the availability of damages under the paragraph above does not make them adequate, because the loss of a catalog and a body of documentation that exist only because we wrote them cannot be measured or restored in money. We may therefore seek an injunction or other equitable relief for that kind of breach, in addition to damages, without having to prove inadequacy of a legal remedy and without posting a bond, to the fullest extent the law allows. Nothing in this paragraph is an admission by either of us as to the amount of any damages.

You will defend, indemnify and hold us harmless against any third-party claim, and any damages, settlements, fines, and reasonable legal costs arising from it, that results from a call you made, an input you submitted, how you used a provider's output, or your breach of these terms. We will tell you promptly about any such claim and will not settle it without your consent, which you will not unreasonably withhold; you will not settle in a way that admits fault on our behalf or imposes any obligation on us without our consent. The limits in this section do not cap what you owe under this paragraph.

We may update these terms as the service changes. The effective date at the top of this page always reflects the current version.

For a change that materially reduces your rights, we will give notice by email to the address on your account at least 14 days before it takes effect. Continuing to use Omnial MCP after that is acceptance. If you do not accept, stop using the service and request a refund of your unused balance. As with any refund, that is a Stripe refund to your original payment method that we arrange by hand, not something that happens on its own; see the refund policy for how it works.

These terms are governed by the laws of the State of Georgia, USA, without regard to its conflict-of-laws rules. The state courts located in Carroll County, Georgia and the United States District Court for the Northern District of Georgia (Newnan Division) have exclusive jurisdiction over any dispute arising out of or relating to these terms or the service, and both of us consent to personal jurisdiction and venue there and waive any objection based on inconvenient forum. Nothing in this paragraph prevents either of us from seeking injunctive relief in any court with jurisdiction to grant it, or you from bringing a claim in small-claims court in your own area.

There is no mandatory arbitration clause and no class-action waiver here.

Talk to us first. Email pixellabsweb3@gmail.com and give us 30 days. Nearly every dispute a service like this produces is a billing question, and most are fixed the same day.

These terms are the whole agreement between us about the service, and they replace anything said before them, on this website or anywhere else, about the same subjects. Our other pages explain how the service works; where one of them and these terms disagree, these terms govern. They can only be changed in two ways: by us updating them as described in changes, or by a document signed by an authorized officer of each of us that says it is changing them. Nothing either of us says or does in the course of using the service changes them, and if we do not insist on the exact terms of this agreement at some point, we can begin insisting again by telling you so.

If your account is closed or these terms end for any reason, the sections on your account, what you may not use this for, what we own, what we do not promise, limitation of liability, and this section all continue to apply. The license in ownership ends the moment your account closes or we revoke it; the restrictions on what you may do with our content do not.

If a provision of these terms is unenforceable, the rest stays in force. Our not enforcing something is not a waiver of it.

Pixel Labs Solutions LLC
pixellabsweb3@gmail.com

Related: the privacy policy covers what we do with your data, the refund policy covers billing and getting your money back, and subprocessors names exactly who your data and your call inputs go to.