Why m18t uses BYOK

BYOK means you bring your own keys: m18t routes AI, social, and CMS calls through your credentials at the provider's rate with no markup, no token resale, no

What you'll learn

  • What "bring your own key" (BYOK) means, and why m18t routes AI, social, and CMS calls through your credentials instead of bundling its own.
  • The cost difference between a bundled-token SaaS and BYOK — who pays whom, and where markup hides.
  • Where your keys go once you hand them over: the encrypted Vault.

The short version

m18t does not sell you AI tokens. It does not sell you social-publishing access. It does not resell CMS hosting. You bring your own keys from the providers you already use — OpenAI, Google (Gemini), Anthropic, Meta, your Strapi instance — and m18t makes the API calls on your behalf using your credentials.

You pay each provider directly, at their published rate. m18t adds nothing to that bill. During beta, m18t itself is free to use — and that's not a promotion that expires into a markup later, it's a consequence of the architecture: there's no pooled token cost for m18t to recover.

If you've used PostHog or a self-hostable tool before, the mental model is familiar: the tool is the workspace; the expensive third-party services stay in your account, on your terms.

A few terms, defined once

  • API — the channel one server uses to talk to another. When m18t publishes a post to a Facebook Page or asks an AI model to draft copy, it sends a request over that provider's API.
  • API key — the credential that proves a request is authorized by your account. It's the thing you generate in a provider's dashboard and paste into m18t.
  • Token (AI billing unit) — AI providers bill per token, roughly a few characters of text. You pay for the tokens you send (your prompt) and the tokens the model sends back (its answer). Image models bill per image instead. The point: the meter runs in your provider account, and you can watch it there.

The bundled-SaaS model, and where it pinches

Most AI-flavoured SaaS tools bundle the cost. You pay a flat monthly fee and get "AI features" or a quota of generations. Under the hood, the platform buys tokens in bulk and resells them to you inside the subscription. That arrangement creates three structural frictions:

  1. Markup. The platform has to protect a margin on compute it bought wholesale. You pay above the raw provider rate, whether you use the AI heavily or barely touch it.
  2. Artificial quotas. To stop heavy users from eroding that margin, the platform caps usage — "200 generations a month," "AI credits," and so on. The cap exists to protect the platform's economics, not yours.
  3. Vendor lag. When a provider ships a new model, you wait for the platform to negotiate pricing and wire it in before you can use it.

None of those are villainy — they're what bundling requires. They're just costs you may not want to carry.

What BYOK changes

Under BYOK, the billing relationship is between you and the provider. m18t is the workspace that orchestrates the calls; it is removed from the money.

  • No markup. You pay OpenAI, Google, or Anthropic exactly their rate. m18t adds zero.
  • Your usage, your dashboard. Every token you spend shows up in your provider's own console. You set the spending caps there.
  • No platform quota on AI. As long as your provider account is funded, m18t doesn't impose an artificial limit on how many drafts or images you generate. (Plan-level limits on brands and connections are a separate thing — that's m18t's own pricing axis, not an AI tax.)
  • New models when the provider ships them. Because you point at the provider directly, you're not waiting on a middleman to resell the latest model.

The honest trade-off: BYOK means setup. You have to create accounts and generate keys before the AI buttons light up — and until a key is in the Vault, AI features stay disabled with a tooltip telling you why. That's a real first-run cost. We think transparent pricing and direct control are worth it, and the setup is one-time per provider. The how-to guides (OpenAI, Gemini, Anthropic) walk each one.

Where your keys live

Handing credentials to any tool deserves scrutiny. In m18t, keys don't sit in plain text. When you save one, it's encrypted at rest with AES-256-GCM in your workspace's Vault, referenced afterward by name rather than by value. The full mechanism — encryption, runtime resolution, rotation — is its own article.

BYOK also fits m18t's shape: it's multi-brand. Each brand can carry its own keys, or several brands can share one. You decide which, per brand.

FAQ

Does m18t make money on my AI usage? No. There's no markup. You pay the provider; m18t isn't in that transaction. m18t's eventual pricing is built on a different axis (brands and connections per account), not on reselling tokens.

Why is the AI button greyed out / disabled? Because no AI key is configured for the active brand. AI features stay disabled, with a tooltip, until a provider key is in the Vault and wired to a connection. Add one via the OpenAI, Gemini, or Anthropic guide.

Which provider should I start with? Any of the three supported text providers works. Gemini carries a "Recommended" badge in the connection picker for fast, low-cost text work; pick whichever you already have an account with. You can wire more than one and choose per use case.

Is it really free during beta? The m18t workspace is free during beta. Your provider bills (OpenAI, Meta, etc.) are separate and always yours — that doesn't change when beta ends.

What if I don't want to use any AI? Then you don't need AI keys at all. The AI-driven surfaces stay disabled and the rest of m18t — content, publishing, CMS push, assets — works without them.

What's next

Was this article helpful?