> ## Documentation Index > Fetch the complete documentation index at: https://docs.humic.dev/llms.txt > Use this file to discover all available pages before exploring further. # Your own OpenAI key > Humic's built-in agent runs on your own OpenAI API key. Humic's built-in agent runs on your own OpenAI API key. OpenAI bills your OpenAI account directly for every agent message you send. Until you add a key, the agent can't answer. You need an **API key** from an OpenAI Platform account with billing set up. A ChatGPT Plus, Pro, Business or Codex subscription doesn't include API access, so it can't be used here. ## Add your key 1. Create a key at [platform.openai.com/api-keys](https://platform.openai.com/api-keys). Make sure the project it belongs to can use GPT-5.6 Luna. 2. In Humic, open **Settings → Agent**. 3. Paste the key under **OpenAI API key** and click **Save and test**. Humic sends one tiny request to OpenAI to check it, and only saves a key that works. Your next agent message uses your key. Everything else stays the same: the same agent, the same tools, and nothing changes in your design system until you click **Approve**. ## Test, replace or remove it * **Test** asks OpenAI again, for example after you added credit. * **Replace** swaps in a new key. Humic tests it before saving. * **Remove** deletes Humic's copy. The key stays active in OpenAI until you revoke it there. If you remove your key or OpenAI stops accepting it, the agent tells you and links to your agent settings. ## How Humic handles your key * It is yours alone. It works in every workspace you belong to, only for your own messages, and nobody else in any workspace can see or use it. * It is stored encrypted. After you save it, Humic only ever shows its last four characters. * Requests are sent to OpenAI with storage turned off, so your design-system content isn't kept in your OpenAI response log. * Humic still limits how many agent messages you can send per minute and per hour, because the agent's tools run on Humic.