
Features
Artificial intelligence has the potential to enhance our work and help us in our writing, thinking, and organization. Over time, Capacities will supercharge more and more workflows with the help of AI. The first step is the Capacities AI Assistant.
At the moment, Capacities supports the following AI features:
Go to Settings > AI Assistant to activate and configure AI in Capacities.
If you use AI features in Capacities, some of your data will be shared with the selected model providers. You can find more information in our Terms and Conditions and our Privacy Policy. We describe the data that is shared with model providers with every feature below. No data will be shared if you don't use any AI features.
If you want to use AI features, you agree to with the terms of use of the model provider you select. You can find their terms on the respective website. Model providers may disallow the use of AI for illegal activity, the generation of hateful or abusive content, the generation of malware, and more.
With your subscription, you get access to the Capacities AI. At the moment, the Capacities AI assistant uses the latest OpenAI GPT models.
With Capacities AI, your data is only processed by model providers, but not stored nor used for training and other purposes.
You get a daily budget for the AI assistant. It is a budget for your Capacities account, not an individual space. You can see the status of your budget in Settings > AI Assistant.
📺 Watch the setup video here.
If you want to use models from other providers or you need to use AI in Capacities beyond the daily budget, you can add an API key to add additional models to the Capacities AI. An API key is a unique credential that authenticates your account with a model provider, enabling direct access to their AI models in other applications.
You may connect your own API keys from a variety of leading LLM providers, giving you full control over the models you use and your usage limits.
You can add API keys from the following providers:
| Provider | API Key Management Link | Model Information |
|---|---|---|
| OpenAI | OpenAI Platform | OpenAI Models |
| Gemini (Google) | Google AI Studio | Gemini Models |
| Anthropic | Anthropic Console | Claude Models |
| Mistral AI | Mistral Platform | Mistral Models |
| xAI | xAI Platform | Grok Models |
| Perplexity | Perplexity AI | Perplexity Models |
API usage with chatbot subscriptions
Some chatbot application memberships, like the ChatGPT Plus subscription, do not include API key usage. You must set up a separate billing method for API usage for your model provider account.
Models available
Only models that produce text output are available for usage within Capacities. If you're missing a model or a provider, please create a ticket in our feedback board.
Training Data
If you want to use your own provider, please review their terms and conditions before generating an API token to understand that provider's policies on training with user data.
Settings > AI Assistant > Custom AI Provider.+ Add new API key.Once a key is added, you can configure the following settings:
How are costs calculated?
Usage costs are calculated based on the public model prices in US dollars per million tokens, and the consumed tokens during usage (input, output and cached input).
When using any AI feature, the Capacities AI Assistant attempts to use an available key in this order:
The AI Assistant model dropdown will display all models made available by your custom keys that are valid for the space you are in. If you have multiple keys from the same provider (e.g., two different OpenAI keys), the Key Name (Alias) will be displayed next to the model name (e.g., "GPT-5 (Work Key)").
You can launch the AI chat in the top left of the application by clicking on the "stars" icon or by using ⌘ + J on Mac or Ctrl + J on Windows.
You can start chatting right away.
You can also choose to select some content from your notes to give the AI Assistant context. The content you select is passed to the AI as a reference.
Currently, the following options are available:
If you select context first, you will get a floating AI chat window when you press Cmd + J. If you want to open it in the side panel, use Cmd + O (Mac) or Ctrl + O (Windows)

If you have added some context, you have access to a range of commands to use with the AI chat. They are most valuable if used with a block selection.
You can trigger the command menu by typing / or +. There are several in-built commands that you can use:
After receiving a response from the AI, you can continue chatting to improve the result or ask follow-up questions.

Once you have achieved your goal, you can select from different actions. You can choose them by clicking on the buttons or by pressing /

You can copy a message, replace your selection with the message, or append it to your active document.
If you want to start the chat again or use a different prompt, go to your original message, click the three dots, and 'rephrase'.
You may want to save your most important or eye-opening AI chats for further reference. This brings you all the functionalities that other objects have:
Hover over the top of the AI chat and click the 'Save as object' button to do this.

Once you've saved the chat, you can open it on its own screen. Here you can add tags, add it to collections, and rename it for easy referencing from other content later.

If you have used any context for the chat, the saved AI chat will be referenced in that object's backlink section. You can also reference the AI chat as you would any other object, via the usual @ or [[]] commands. This helps truly integrate the AI chat into the rest of your knowledge work.

These saved chats are also fully searchable via the full-text search. The search results will take you to the message that holds the search term which allows you to quickly find content you know you've saved.

The AI assistant is especially helpful if you're on the go. It is fully supported in our mobile app except for the use of commands.
To launch the AI assistant in the mobile app, you can select a context (e.g., a block selection or by opening the corresponding note), then press the "AI Assistant" button. Saving your favorite AI chats as objects is supported on mobile too.
The AI panel is a quick way to perform tasks such as translation, rewriting, or summarization. The panel floats over the top of the text.
To open it, select text and click on the 'sparkle' icon that appears in the context menu.
The panel provides quick actions like Summarize, Translate, Improve Writing, and more. You can see the full list when you open the panel.

If you don't want to launch the AI chat, there is a quick and convenient way to ask a question.
Type your question into a new block in your content and then type "/Ask AI"

You will get a response to that question below.

Tags are an important organizational structure in Capacities, and AI can help you use them.
AI identifies relevant tags from your existing tags and assigns them to that object. You can edit the selection if needed afterwards.
Data protection
Your current object's properties, and the beginning of your notes in that object will be sent as a context to the AI.
This option appears anywhere you see tags for an object.

Here's a video where you can see this in action:
Similar to the AI Auto-tagging, you can also use AI to select collections for your objects. Simply click on the AI button in the collection picker. The AI will then select the most relevant collections for your object. It will not create new collections but only select existing ones.
Data protection
Your current object's properties, and the beginning of your notes in that object will be sent as a context to the AI.

You can also use AI to auto-fill your properties. This allows you to quickly insert a fact summary or use the AI's knowledge so you don't have to research it.
This works for the following properties:
Data protection
Your current object's properties, and the beginning of your notes in that object will be sent as a context to the AI.
You can activate the AI property auto-fill in the property settings. You can also choose to add more instructions in the property settings to get more accurate results.
It's important that you first add content to some other properties (at least the title) so the AI has some context to fill your property.
After activating auto-fill for your property, you can simply click on the button that appears in your property field when you hover over it.
If you have several properties with AI autofill, hover over the title of the object and click 'Fill properties' to fill them all at once.

If you're not happy with the result, you can add additional instructions to your property auto-fill in the property settings.
Additional instructions are added to the context that is sent to the AI. This allows you to give the AI more information about what you want to achieve.
Here are some examples:
You might need to play around with different options. It helps to be very specific. You are also welcome to share your best practices in our community.
We envision AI to be used in every layer of Capacities to make your workflows simpler, faster, and more powerful. Read more about our AI roadmap.
If you have any ideas, problems, or feature requests, please let us know on the feedback board.
If you encounter an error message while using the AI assistant, the root cause is usually related to your API key, budget, or access permissions for the selected model provider.
| Error Message / Category | Potential Cause(s) | How to Fix |
|---|---|---|
| "Invalid API Key" | The API key entered is incorrect, contains extra spaces, or has been revoked by the provider. | Double-check the key for typos. Try generating a new key from the provider's platform and try again. |
| "Insufficient Funds" / "Rate Limit Exceeded" | The key's account has run out of credit, or the key is hitting a provider-set spending limit. | Top up your balance or check your billing setup on the provider's platform (e.g., OpenAI Billing). Increase the Max Budget setting for the key in Capacities. |
| "Model Not Found" / "Permission Denied" | The specific model you selected (e.g., a newer GPT-4 variant) is not enabled for your API key, or your access tier does not include it. | Check the provider's documentation and your account settings (e.g., OpenAI Rate Limits) to confirm model access. Select a different, more widely available model. |
| "Quota Exceeded" | Your usage has hit a time-based limit (e.g., tokens per minute) set by the provider for the model. | Wait a few minutes and try again. For continuous high usage, check the provider's platform to see if you can request a rate limit increase. |
| "Internal Server Error" | A temporary issue on the model provider's side. | Try again in a few minutes. Check the provider's status page (if available) for known outages. |
| "Browser extensions blocking the AI assistant" | Security or ad-blocking extensions may interfere with the request. | Try disabling browser extensions or use the dedicated Capacities desktop application. |
Settings > AI Assistant where you can activate or hide all AI features. Additionally, you can bring your own API key from a model provider if you wish to use other models or use the AI assistant without any budget limits.Settings > AI Assistant > Custom AI Provider.Settings > AI Assistant > Preferences. Capacities' AI uses OpenAI, but by toggling this option on, you are replacing Capacities AI with your European model provider.