Many tools add AI features by bundling them into a subscription. KanbanWP takes a different approach: you connect your own account with an AI provider, and you pay that provider directly for what you use. KanbanWP does not resell AI, add a markup or need a payment gateway.
This post explains how that works, what the plugin uses AI for, and the controls that keep usage under your management.
Supported providers
You choose one provider in the plugin settings and paste an API key:
- OpenAI
- Anthropic
- Google Gemini
- A custom provider, which needs a base URL and a model name as well as a key
Each built-in provider has a default model, and you can override it with a model of your choosing. Setup steps are in providers and keys.
Until a key is saved, AI features are simply unavailable. Nothing is sent anywhere.
What AI does in the plugin
The plugin includes a set of focused features, each with its own prompt that you can add instructions to:
- Text to tasks: paste rough notes and get structured task suggestions.
- Task breakdown: split a card into subtasks.
- Task summary: summarize one task.
- Post outline and post draft: for tasks linked to editorial work.
- Post SEO: suggest a title, meta description, excerpt and keywords.
- Board summaries: weekly and daily summaries of a board.
- Generate board: draft a board structure.
- Client updates: friendly or professional progress notes you can adapt.
Treat all output as a draft. The model sees what you send it, and it can be wrong, so review before you publish or send anything to a client.
Who can use it
AI is not open to every logged-in user. Plugin administrators can always use it. For others, you choose which WordPress roles are allowed in the settings; by default only administrators are. That keeps costs predictable on sites with many editors or contributors.
Monthly token caps
Tokens are the unit providers bill by. KanbanWP lets you set a monthly token cap per user. When a person reaches the cap, further AI requests are refused with a message that their monthly budget has been reached, until the next month. A cap of zero means unlimited.
A few practical ways to pick a number:
- Start with a modest cap, such as a figure that covers a few dozen requests per person.
- Check usage after two weeks.
- Raise the cap for people who clearly benefit, and leave others low.
Also set spending limits in your provider’s dashboard, as a second safeguard that does not depend on any plugin.
Usage logging
Every successful request is written to a usage table with the user, board, feature, provider, model, and input and output token counts. You can review this in the plugin, so you can see which features and which people account for your spend. See usage and limits.
The log records counts, not the text of your prompts or answers.
How your key is handled
The key is used on your server when the plugin calls the provider. It is not sent to the browser, it is not written to the usage log, and the plugin redacts anything that looks like a key from provider error messages before showing them. Still, treat it like any credential: create a key dedicated to this site, restrict it in the provider dashboard if possible, and rotate it if you suspect exposure.
What leaves your site
Board data stays in your own database. The only thing sent to a third party is what you explicitly submit through an AI feature, and it goes to the provider you chose. Check that provider’s data handling terms before sending anything sensitive.
A sensible rollout
- Connect a key and test with a throwaway board.
- Restrict access to administrators first.
- Set a monthly cap.
- Try text to tasks and task breakdown, which tend to save the most time.
- Review the usage log after a couple of weeks and adjust.
If you would rather let an external AI agent work with your boards instead, read AI agents and MCP for WordPress boards. The full reference for AI settings starts at the AI documentation.
Wondering which AI features save the most time? See best AI task managers in 2027 and the overview of AI project management.