Marek runs a 14-person agency in Poznań. On Monday, he put the AI rollout on hold after calculating the subscription cost for the whole team. He returned to the idea when he discovered a different billing model: using the company’s own AI key, paying the provider directly and receiving a bill based on actual usage. In practice, your own AI key is an API key - credentials that connect Task Manager to your account with the model provider.
Using your own AI key removes the intermediary from billing
Using your own AI key means that OpenAI, Google Gemini, Anthropic Claude or xAI Grok bills you directly for use of the model. Task Manager does not add a markup to requests or charge a per-user subscription for AI.
The software licence and model usage are billed separately. You buy the software licence once for a specific number of seats, while model usage is billed according to the chosen provider’s pricing. The Standard option in the Classic plan for a team of up to 20 people costs 1 490 zł and includes 12 months of updates. After that, you can purchase an annual renewal or continue using your current version without further licence fees.
There is no single model provider bill that would apply to every company. The cost depends on the chosen model, the number of requests and the response length. However, the charges are billed to the company account created with the provider, not by an intermediary.
AI remains optional. Without a provider account and API key, Task Manager still supports your team’s day-to-day work, while the AI buttons simply do not appear. You can read more about this billing model on the Your model, your key page, while the difference between a one-off licence and a subscription is shown in the cost comparison for a 12-person team.
How to connect an API key in five steps
You connect the API key in the AI tab, which is available from the application settings. An administrator configures AI by selecting a provider, pasting in the key, retrieving the models from the account and saving the settings, after which the application tests the connection.
- Click the cog icon in the top-left corner and open the “AI” tab.
- In the “AI Provider” field, select OpenAI, Google Gemini, Anthropic Claude or xAI Grok.
- Paste the key into the “API Key” field. You can generate it in the provider’s console:
platform.openai.com/api-keys,aistudio.google.com/apikey,console.anthropic.com/settings/keysorconsole.x.ai. - Click the refresh icon (↻) next to the model list - the application will retrieve the latest list of models from your provider account. Select a model, or if you are just getting started, choose one marked “(recommended)”.
- Click “Save”. Task Manager will test the connection to the model and display any warnings.
The key is encrypted in the database and cannot be read in full on the screen after it has been saved. If you need to change it, paste in a new one.
The server must support PHP 8.1 or later and MySQL or MariaDB, while the curl extension is required for connections to the AI provider. Standard shared hosting is sufficient. You can find the exact field layout in the step-by-step AI settings guide.
Providers differ in their models and semantic search support
Your choice of provider determines where the charges are billed and which models your account offers. OpenAI and Google Gemini support semantic search; Task Manager retrieves the list of available models from your account with each of the four providers.
| Provider | Where to generate the key | What you need to know |
|---|---|---|
| OpenAI | platform.openai.com/api-keys | Supports semantic search, which is enabled by default for this provider. |
| Google Gemini | aistudio.google.com/apikey | Supports semantic search, which is enabled by default for this provider. |
| Anthropic Claude | console.anthropic.com/settings/keys | Semantic search requires OpenAI or Google Gemini. |
| xAI Grok | console.x.ai | Semantic search requires OpenAI or Google Gemini. |
Task Manager retrieves the latest list of models from your account, so you do not need to enter their names manually. Semantic search finds content by meaning rather than only matching identical words. This feature requires OpenAI or Google Gemini.
If you decide to switch providers after a few months, return to the same tab. Select another provider, paste in the new key and retrieve the model list again. You do not need to reinstall Task Manager.
Set four safeguards before granting access
The most effective way to limit API usage is to do so before enabling AI for the whole team. Set the response length, file summarisation method, Automations limit and knowledge base update frequency.
1. Reduce the maximum response length
The “Maximum response length” field accepts values from 1024 to 16384 tokens, with a default setting of 4096. A token is a fragment of text counted by the model provider. A higher value allows longer responses but increases usage.
To begin with, do not increase this value just because the field allows it. Marek keeps the default 4096 and only increases it when the team needs longer responses.
2. Keep attachment summaries in Manual mode
“AI attachment summaries” are set to Manual mode by default. The model only receives the file after you click “Create summary”, so simply adding a document does not trigger an automatic request.
Each summary means one API call per file. Automatic mode runs in the background and requires a cron job, which is a recurring task run by the server.
3. Limit AI in Automations
The “Daily limit for AI requests from Automations” is set to 200 requests per day by default. Today’s usage is shown next to the field, and entering 0 disables the limit.
Once the limit is reached, Recipes that use AI are skipped for the rest of the day, and this is recorded in the history. The counter resets at midnight. The limit applies only to Automations, not to all AI features in the application. The Automations module itself is available with Pro.
4. Update the knowledge base deliberately
The AI knowledge base, available with Pro, is updated once a day in the background by default. It allows the AI Assistant to find answers in company Notes and Files without you having to specify a source.
Notes and Files are added to the index without being summarised by the model, so indexing them does not use AI requests. Tasks and Chats work differently: they are summarised by the model. If you have a very large amount of this content, you can set “Automatically update the knowledge base” to “No” and run the process using the “Update AI knowledge base” button.
The capacity bar below the sources shows how much knowledge fits into the context. If you select too much material, the AI Assistant will indicate what it has not included.
You set AI access separately for each person
Each user can be given three independent permissions: AI in Tasks, AI in Chats and access to the AI Assistant. A fourth permission allows the AI Assistant to retrieve content from a specified website and is disabled by default.
On Monday, Ola, a coordinator at Marek’s 14-person agency, receives all three basic permissions. Two warehouse employees receive none, so the AI buttons simply do not appear in their Tasks and Chats views. They continue to use Tasks, Chats and the other modules as normal.
You set permissions when adding or editing a user in the Users window. The global administrator has all AI permissions by default. A detailed breakdown is available on the Users and Roles in Task Manager page.
You can start the test with three people, see how they work and only then expand access. You do not need to enable AI for all 14 accounts at once.
Your own key does not mean AI runs locally
The request and the selected context are sent to your chosen model provider. Using your own key means having your own account, being billed directly and working without an intermediary, but it does not mean the model runs locally.
The boundary between the company server and the model provider is particularly important for Mail. The “Allow mail to be used as an AI knowledge source” option is disabled by default. Once it has been deliberately enabled, each user can only select their own mailbox, and even the administrator cannot view anyone else’s correspondence. Mail and the option to use it as a source are available with Pro.
Correspondence is not added to the knowledge base indexed in the background. If the administrator withdraws permission to use Mail, Task Manager also stops using any Mail sources selected previously.
Threads in the AI Assistant are private. Other users and the administrator cannot see them. The AI Assistant only uses materials that the person asking the question has permission to access.
The message identifies the cause of the AI issue
Task Manager shows the reason when the model does not respond. Usually, you need to correct the key, change the model, refresh the list or wait for the provider’s limit to reset.
“Error: invalid API key” means you need to check the key you entered and the active payment method or limits on your provider account. If the model list is empty, click refresh. If that does not help, check the key and the server’s internet connection.
“The model does not support chat mode” appears when the selected option is not a chat model. Return to the settings and select a different model from the downloaded list.
“Request limit exceeded (rate limit)” relates to your provider account. You can wait or upgrade to a higher plan with the provider. This is separate from the daily limit for AI requests from Automations set in Task Manager.
Tomorrow, create an account with one provider, enter the key, keep the response limit at 4096 and enable AI for three people for one week. After a week, check the bill in the provider’s dashboard and only then decide who else needs access. Launch the Demo to see Task Manager in action.
Frequently asked questions
How much does AI cost in Task Manager?
Task Manager does not charge for AI or add a markup to requests. You pay the provider that issued your API key directly, based on its pricing and your actual usage, while the Task Manager licence is a one-off purchase with no per-user subscription.
Is data sent to OpenAI, Google Gemini, Anthropic Claude or xAI Grok?
Yes, the request and the selected context are sent to your chosen provider using your key. That is why Mail is disabled as a knowledge source by default, and the AI Assistant can only access materials available to the individual user.
Does Task Manager work without AI?
Yes, Task Manager works without AI, and an account with a model provider is optional. Without a configured key, you can still use Tasks, Chats, Calendar, Notes and other features.
Can I switch from OpenAI to Google Gemini, Anthropic Claude or xAI Grok later?
Yes, you can change providers in the same AI tab. Select the new provider, paste in its key, retrieve the list of models and save the settings.
How can I limit the bill for my company’s own AI key?
Reduce the maximum response length, leave attachment summaries in Manual mode and set a limit on AI requests from Automations. Start by granting access to people with a specific use for AI rather than enabling it for the entire team straight away.
Sources
The Task Manager Team - we write about teamwork, AI in business and keeping tasks organised, using examples from Task Manager, software installed on your own server. Task Manager has been developed by Blue Creation Group (Piotr Łącz) since 2013 and is used by organisations including Neonet, the Military University of Technology and Polpharma - read customer reviews. We check our articles against the user guide and the latest version of the software.
