In **AI** you decide, provider by provider (**OpenAI**, **Claude (Anthropic)**, **Gemini (Google)** and **DeepSeek**), whether your customers use **AI managed by the Platform** (usage is deducted from your wallet, Manual 39) or your own key for that provider (BYOK: "Bring Your Own Key"). The screen has five tabs: **Models to use**, **Keys**, **Prices**, **Negotiated** and **Margin**. ## 4.1 Models to use: which models you offer your customers It's the first step for everything else: only the models you enable here, provider by provider, become available afterward in Keys, Prices, Negotiated, and in your customers' agent model selector (Manual 23). Each provider group carries a badge with your real status with it: - **"With your key"**: you use your own key for that provider (section 4.2, below). - **"With Platform's key"**: you don't have your own key for that provider — your customers use the Platform's key and usage is deducted from your wallet at your prices (Manual 39, section 1); the provider's group itself repeats the notice: "Usage is debited from your wallet at your prices". - **"No key configured"**: neither you nor the Platform have a key for that provider ready for your account yet — you can still turn its models on or off from here regardless. **Requirements:** only the Owner role can turn models on or off; the rest of your team sees the screen read-only. Turning a model on is instant — just another toggle, no confirmation. Turning it off ALWAYS asks for confirmation, with the real effect stated upfront: *"You're about to turn off [model] from [provider]. Your customers will stop seeing it on their agents."* ### Procedure: asking support to turn on a model you don't see 1. At the bottom of "Models to use", tap **"Request from support"** — anyone on your team can do this, no Owner permission needed. 2. The **"New ticket"** modal opens (the same one from your own support, Manual 42, section 3 "My support") with the **Subject** already filled in: "AI model activation". Fill in the **Description** with the model you need. 3. Tap **Create ticket**. **Result:** the button is replaced by the notice "Ticket sent. The support team will review it shortly.", and the ticket becomes visible on your own support (Manual 42, section 3) like any other. ## 4.2 Keys: one or several per provider You can have **several keys for the same provider** — for example, a "Premium customers account" and a "Budget account" for OpenAI — and split them among your customers as needed. Each provider card lists your keys with the **Name** you gave it and a masked version of the real key (to recognize it without exposing it), and offers **Model** (sets that key's default model) and a trash icon to delete it. ### Procedure: adding a key **Requirements:** Owner role — same as turning models on/off (section 4.1), the rest of your team sees Keys read-only. Available for **OpenAI, Claude (Anthropic) and Gemini (Google)** — DeepSeek doesn't offer this button for now (see the warning below). 1. On the provider's card, tap **"+ Add key"**. 2. Type the **Name** — to identify this key if you have several from the same provider. 3. Paste the **API key**. 4. Tap **Save key**. **Result:** the key shows up on its provider's card, ready to assign to your customers. If you try to delete a key still assigned to a customer, the console warns you and doesn't delete it until you remove that assignment first. Each key's **Model** button opens a selector — you no longer type the model by hand: you can only choose among the models you've enabled for that provider in **Models to use** (section 4.1), with a "No default model" option if you'd rather not set one. If the model you're looking for doesn't show up in the list, the selector itself reminds you: "Don't see a model? Enable it in Models to use". > **Warning — DeepSeek: postponed for now** > > The **DeepSeek** card carries the **"Postponed"** badge and doesn't offer the "+ Add key" button — unlike > OpenAI, Claude and Gemini, you can't give it your own key yet. The card's own text explains it: "DeepSeek > is postponed for now: we'll enable it once EU-hosted deployments are available. In the meantime, keys for > this provider can't be added." This doesn't affect **"Models to use"** (section 4.1): its models still > show up there and you can turn them on or off just like the rest of the providers' — the only thing > blocked is your own key. ## 4.3 Customer assignment: which key and model each one uses In **Customer assignment**, search for the customer by name and, for every provider where you have any key of your own, choose the **Key** and the **Model** that specific customer should use, and tap **Save**. A provider with no key of your own shows "You don't have any key of your own for this provider yet" instead of the selector. > **Note — Example: the same key, two customers, two different models** > > Nothing forces you to use one key for a single customer or a single model. You can, for example, have 4 > keys spread across your providers and assign the SAME OpenAI key to two different customers, each with > their own model — for example, one customer on `gpt-4o` and another on `gpt-4o-mini`, both on the same > "Premium customers account". ## 4.4 Prices: what you charge your customers Sets the price you charge your customers per model. Next to it, under **"What the Platform charges you"**, you have your own reference list — the price your contract already gives you, your margin reference, never editable from here. ### Procedure: setting a resale price 1. On "Your resale list", tap **"+ New price"**. 2. Choose the **Provider** and the **Model** — both are selectors: Model is populated with the models you've enabled for that provider in **Models to use** (section 4.1); if you don't see the one you're looking for, the "Don't see a model? Enable it in Models to use" hint takes you straight to that tab — and the **Currency** (EUR by default). 3. Type the **Input price** and the **Output price**, both per 1M tokens. 4. Tap **Continue**. 5. Review the summary — *"You're going to set €X/€Y per 1M tokens for [provider] · [model], effective now"* — and tap **Confirm**. **Result:** the price is effective from that moment and shows up on "Your resale list". Each row's **"History"** link opens the previous versions with their effective date. ## 4.5 Negotiated: a special price for a specific customer If you agree on different terms with a particular customer (by volume, by contract…), set a **negotiated price** just for them — separate from your general resale list — with a **note** documenting the agreement. ### Procedure: negotiating a price 1. Search for the customer and tap **"+ Negotiate price"**. 2. Choose **Provider**, **Model** and **Currency** — Provider and Model are selectors populated by what you've enabled in **Models to use** (section 4.1), with the same "Don't see a model? Enable it in Models to use" hint if you're missing one; the Currency must match THAT customer's contract currency for the price to be able to apply. 3. Type the **Input price** and the **Output price** (per 1M tokens). 4. Type the **Note** — the agreement's reference; it's **required**, you can't continue without it. 5. Tap **Continue** and then **Confirm**. **Result:** that customer's negotiated-prices table shows the provider, model, both prices, and your note, with its own **"History"** link. > **Note — If you choose the wrong currency** > > This form doesn't validate against the customer's real contract: if the currency doesn't match their > contract's, the negotiated price simply never ends up applying (it never gives an error) — check the > customer's currency beforehand. ## 4.6 Margin: what you bill versus what the Platform debits you **Margin** compares, for the **Period** and **Grouping** (Model or Customer) you choose: what you **Billed** your customers, what the **Platform debits you**, **Your margin**, and the **Margin %** — all in your contract currency. > **Warning — You never see the Platform's real cost here — only what it debits you** > > This view shows what YOU bill and what the Platform charges YOU — never the real cost to the Platform of > providing the service. If a customer uses your own key (BYOK), that row carries the **"With your own > key"** badge and "Platform debits you" stays at €0 — because that usage never goes through the Platform: > it's paid by your own AI provider, with your own OpenAI/Claude/Gemini/DeepSeek billing, outside this > console. If some usage doesn't have a resale price set yet (section 4.4), the table flags it with a **"Fallback"** notice — while you don't set that price, that usage still gets billed to you regardless, but you can't see it broken down as your own margin. ## 4.7 Allowing your customers their own key (different from YOUR own key) This is one more tab on this screen, **"Customers with their own key"** — don't confuse it with "Keys" (section 4.2, above): there you configure YOUR own key to resell it to your customers through your wallet. Here you decide whether your customers can configure their own **completely independent** key — without going through either your wallet or your key, exactly like the "Bring your own AI provider" tab on their own Platform (Manual 76). The **"Allow my customers to use their own AI provider key"** switch is on this tab, with a plain explanation and the date of the last change (or "Not changed yet" if you've never touched it). Only the **Owner** role can change it; the rest of your team sees it read-only. ### Procedure: turning your customers' own key on or off 1. In **AI → Customers with their own key**, click the switch. 2. If you're **turning it off**, confirm in the warning window: *"Customers who already have their own key saved will stop using it immediately and go back to the platform's provider. Their saved key isn't deleted: they can use it again if you turn this setting back on."* **Result:** turning it on does NOT switch the feature on for all your customers at once — it only enables it as an option. The EFFECTIVE activation, customer by customer, is done afterward on their plan copy (section 4.8, or ask support to activate it if your console doesn't let you edit it). Turning it off, however, DOES have an immediate effect on every customer who already had it configured. ## 4.8 AI breakdown by customer On your **Customers** screen (Manual 38), each row can be expanded with **"AI breakdown"**: the same Platform / Own key / Fallback split you see aggregated in "Margin" (section 4.6), but for ONE specific customer. Next to the breakdown, an indicator tells you whether that customer has the **"Bring your own AI provider"** capability turned on — the one you enabled with the switch in the previous section. If you need to **turn it on or off for a specific customer**, open them and edit their plan copy — you'll find the "Bring your own AI provider" row there, just like the rest of their capabilities (same mechanism as Manual 38). > **Note — "Own key" in this breakdown is ALWAYS the customer's own** > > Don't confuse it with the "With your own key" badge in "Margin" (section 4.6): that row is about YOUR key (the one you resell them). This per-customer breakdown, instead, splits what THAT SPECIFIC customer consumes via Platform / their own key (theirs, configured on their own Platform) / Fallback — two different things that can coexist for the same customer.