Bring Your Own Key (BYOK) for AI: Get and Use an API Key

“Bring your own key” (BYOK) has two meanings. In cloud security it means supplying your own encryption keys. In AI tools, which this guide covers, it means the tool calls an AI model with an API key from your own provider account, and the provider bills you directly.

KanbanWP works this way. This guide explains the trade-offs of BYOK AI, shows how to get an OpenAI, Claude or Gemini API key, and covers what the plugin does with your key once you add it.

What bring your own key means for AI tools

With BYOK, the software does not resell AI. You create an account with a provider such as OpenAI, Anthropic or Google, create an API key, and paste it into the tool. Each request goes to the provider on your account, and you pay the provider’s per-token price.

BYOK has clear benefits:

• No markup or AI seat fee. You pay only for the tokens your team uses.
• Your choice of provider and model. You can switch providers or pick a cheaper model for simple jobs.
• Your own data terms. The provider’s API terms apply to your requests, not a reseller’s.

BYOK also has costs:

• Setup work. Someone has to create an account, add billing and manage the key.
• Separate bills. AI usage appears on the provider’s invoice, not with your other software.
• Key security. A leaked key lets others spend on your account until you revoke it.

How to get an API key

Each provider issues keys from its own developer console. Chat subscriptions do not include API keys: OpenAI bills API usage separately from ChatGPT, and Anthropic bills the Claude API separately from paid Claude plans. Anthropic’s article adds that Claude Max and Team plans include monthly API credits.

How to get an OpenAI API key

1. Sign in to the OpenAI platform and open the API keys page.
2. Create a new secret key and give it a name, such as “KanbanWP on example.com”.
3. Copy the key straight away. OpenAI shows the full key only when you create it, so if you lose it you create a new one.
4. Add a payment method or prepaid credits in the API billing settings.

How to get a Claude API key (Anthropic)

Anthropic’s steps to get a Claude API key are:

1. Sign in to the Claude Console at platform.claude.com, or create an account.
2. Open the API keys page.
3. Click Create key, name it, choose an expiration and set the linked account. You can also scope the key to a workspace.
4. Copy the key, which starts with sk-ant-. The Console shows it only once.

Anthropic recommends a service account key for anything shared, such as a website, because a personal key stops working if you leave the organization. Spend limits are on the Console’s billing page.

How to get a Gemini API key

Google’s guide to Gemini API keys gives these steps:

1. Open the API keys page in Google AI Studio.
2. Create a key. Each key belongs to a Google Cloud project, and new users get a default project and key when they accept the terms of service.
3. Copy the key and keep it out of web pages and public code.
4. Set billing alerts in the Google Cloud console.

The Gemini API has a free tier. Google may use free-tier content to improve its products, so use a paid tier for client or confidential work.

Add your key to KanbanWP

Go to KanbanWP > Settings > AI and fill in these fields:

1. AI provider: OpenAI, Anthropic, Google Gemini, or Custom. Custom covers any service that accepts OpenAI-style chat completion requests, such as OpenRouter, Groq or Mistral, and it also needs a base URL and a model name.
2. API key: paste the key. KanbanWP stores it encrypted with your site’s WordPress security salt. After you save, the field stays empty and shows that a key is saved. Leave it empty later to keep the same key.
3. Model: leave it empty to use the provider’s default, or enter a model name. From version 1.0.1 the defaults are gpt-6-luna for OpenAI, claude-haiku-5-5 for Anthropic and gemini-3.6-flash for Gemini. Providers retire models over time. Version 1.0.0 defaulted to gemini-2.0-flash. Google’s deprecation table lists its shutdown on June 1, 2026, so on 1.0.0, update the plugin or enter a current model name.
4. Test connection: sends a tiny request with the values in the form and reports whether it worked and which model answered.

Until a key is saved, every AI feature returns a “not configured” message and no board data goes to a provider. The full steps are in providers and keys.

What the plugin does with your key

KanbanWP handles the key on the server:

• The key is saved in your WordPress database with the plugin’s other settings.
• It is used only inside server-side requests from your site to the provider.
• It is never sent to the browser, and the settings page does not show it again after you save it.
• It is never written to the usage log.
• If the provider returns an error, the plugin removes anything that looks like a key from the message before showing it.

Because the key sits in the database, treat database backups and exports as sensitive. Create a separate key for each site, set a spending limit in the provider’s dashboard, and replace the key if you think it has leaked.

What the key is used for

Each AI feature in KanbanWP has a fixed core prompt. On the AI Prompts tab you can add instructions for all features, such as your brand voice, and extra instructions for each feature. The features are:

• Text to tasks: turns pasted notes into tasks with priorities, labels and due dates.
• Task breakdown: suggests subtasks for one task.
• Task summary: summarizes one task.
• Post outline, post draft and post SEO: write a blog post outline, a draft or SEO fields from a task, or from the WordPress post linked to it.
• Board summaries: a daily standup or a weekly progress report for one board.
• Generate board: creates a board with stages and starter tasks from a goal.
• Client update: a friendly or professional update that leaves out internal tasks and comments.
• Command: runs short instructions such as “move” or “assign” on a board.

In version 1.0.1 these features are reached through the REST API, and the board screen has no AI buttons yet. Treat every result as a draft and review it before it goes to a client.

Who can use AI on your key

AI is not open to every logged-in user. Users who manage KanbanWP can always use it. For everyone else, you tick the WordPress roles that are allowed, and by default only Administrators are. A user also needs edit access to the board a feature works on.

Monthly token caps

Providers bill by the token. The Monthly token cap setting is one number that applies to each user separately. When a user’s total for the calendar month reaches it, their next AI request is refused with a message that their monthly AI token budget has been reached. A cap of 0 means no limit.

To pick a number:

1. Start with a modest cap and check usage after two weeks.
2. Raise it if people hit it during normal work.
3. Keep a spending limit in the provider’s dashboard as well, since it works even if plugin settings change.

The usage log

Every successful AI request is saved with the user, board, feature, provider, model, and input and output token counts. The AI Usage screen shows each user’s tokens this month against the cap, and the 100 most recent calls. The log keeps counts only, not the text of prompts or answers. See usage and limits.

What leaves your site

Board data stays in your WordPress database. When someone runs an AI feature, only the text that feature needs, such as the pasted notes, the task details or a linked post, goes to the provider you chose. Check that provider’s data terms before you send client material.

A short rollout plan

1. Create a dedicated key with a spending limit.
2. Add it under KanbanWP > Settings > AI and run Test connection.
3. Keep AI limited to Administrators at first, and set a monthly cap.
4. Try text to tasks and task breakdown on a test board.
5. Read the AI Usage screen after two weeks, then open AI to more roles.

If you want an outside agent such as Claude to work on your boards instead, read WordPress MCP servers and AI agents. To compare KanbanWP with hosted tools that bundle AI into seats, see the best AI project management tools for 2027, and for the bigger picture, read AI for project management.