Costs

XIV MCP itself is free. The AI app you connect it to usually is not.

AI apps measure use in tokens: small pieces of text, roughly a word each, that the model reads and writes. These have nothing to do with XIV MCP's access token, or with tokens in the game.

• Every request uses tokens. Each time your assistant reads the game or acts in it, the AI model handles a request. Controlling the game step by step uses many: every tool call is a request, and a dungeon farm driven step by step can make hundreds of them. Run as a background job, the same farm costs a few requests (see below).
• Free plans run out quickly. Claude, ChatGPT, GitHub Copilot and Cursor include a small amount of use on their free plans. For regular use you need a paid plan.
• API keys bill every token. There is no limit unless you set one with the provider.
• LM Studio is the free option. Models on your own PC cost nothing per request. They do take a lot of memory and share the graphics card with the game. LM Studio can also use cloud models, which are billed like any other. See LM Studio.

XIV MCP cannot see how many tokens your assistant uses, so check the usage page of your plan.

Using fewer tokensCopy Anchor Link

• Turn off modules you do not use. A module that is switched off disappears from your assistant's tool list, so every request is shorter. See Turning a module off.
• Use background jobs for long tasks. A job runs its steps in the game by itself, without a request per step. Step by step, the assistant pays for every call, and for every time it checks whether something has finished. See Step by step or as a job.
• Start new chats. Long chats send the whole history with every request.