> ## Documentation Index
> Fetch the complete documentation index at: https://workllm.mintlify.site/llms.txt
> Use this file to discover all available pages before exploring further.

# How WorkLLM credits work: costs, usage, and tips

## **Overview**

WorkLLM gives your team one workspace for chat, memory, agents, and connected apps. It also gives you access to more than 200 AI models. Each of these costs a different amount to run. A short question to a light model costs very little. A long research task on a premium model costs much more.

Credits are the single unit WorkLLM uses to measure all of that work. They let you see what you use, compare your options, and plan your monthly usage. This matters most for agencies and consulting teams, where several people share one workspace and one pool of credits.

This article explains how credits work in practice. It covers the following topics.

* What a credit is, and what uses credits and what does not.
* What affects the cost of a task, and what typical tasks cost.
* How many questions your credits cover, with examples for an agency and a consultant.
* How to see your usage, make your credits go further, and buy more.
* How teams share credits, and what to do if your credits drop faster than expected.

## **How credits work**

### **What is a credit**

A credit is how WorkLLM measures the AI work you use. We convert the model provider's token cost into credits. We also include storage, infrastructure, and platform costs. So what you spend reflects what the work costs to run.

A credit is not a token. It is not a question or a message. It is also not the same as an OpenAI or Anthropic credit.

Credits are not a flat price per question. A premium model uses more credits than a lighter model for the same answer. Every plan includes a monthly pool of credits. Everyone in your workspace shares that pool.

### **What uses credits**

Credits are used whenever you or your team ask WorkLLM to do something.

| Name | Explanation | Typical credit cost |
| :- | :- | :- |
| Chat Answers | Every message you send to WorkLLM uses credits. The amount depends on the model you pick, the length of your question and answer, and the length of the conversation. | 1-page answer: Lighter models: 3–6 credits; Medium models: 5–10 credits; Higher-end models: 8–15 credits |
| Web Search | When WorkLLM searches the web for you, it adds a small extra charge on top of the answer. | About 2 credits |
| Deep Research | Deep research works across many sources in several steps. It uses noticeably more credits than a normal question. | About 10–30 credits |
| Documents and Files | Saving a file to memory uses credits once to process it. A file attached to a chat is read together with your question. | About 5–20 credits, depending on the file and the work required |
| Image Generation | Creating or editing an image uses more credits than a text answer. The cost depends on the image model. | About 10–30 credits |
| Agent Creation | Creating an agent uses credits. When you describe what you want, WorkLLM does model work to build the agent and its steps. | About 10–30 credits |
| Agent Execution | Each time an agent runs, it uses credits for the model work in every step. An agent with more steps uses more credits. You can see the cost of every run. | About 5–30 credits per run, depending on the number of steps and models used |
| Connected Apps | Credits are used when you connect an app for the first time. Credits are also used each time you ask WorkLLM something in a connected app, such as finding an email in Gmail, reading a Google Sheet, or posting a message in Slack. | 10 credits for the initial connection, one time. After that, usage is based on the task |

### **What does not use credits**

Some things do not use WorkLLM credits.

* **Your own API key (BYOK).** When you pick a model that runs on your own API key, the model itself does not use WorkLLM credits. You pay your provider directly for those tokens. To set up your key, go to Workspace Settings, then BYOK.
* **Searching your saved memory.** Looking up a file you saved to memory has no separate charge. You pay for the answer only.
* **Failed requests.** A request that fails should not use credits. If you were charged for one, email [hello@workllm.io](mailto:hello@workllm.io) and we will refund the credits.

Some things still use credits when you use your own key. These are memory, web search, and embeddings. WorkLLM never switches between your key and your credits on its own. The model you pick decides which one pays.

## **What things cost**

### **What affects the cost of a task**

Six things affect the cost of a task.

* **The model.** Heavier models cost more than lighter ones. The model picker shows a cost level for each model, from one dollar sign to four dollar signs. You can sort the list by cost.
* **The Recommended setting.** On the default Recommended setting, WorkLLM picks the best model for each task based on cost and quality. It may pick a stronger model for a complex request. If you choose Recommended Low, it only uses lower-cost models. Recommended Medium and Recommended High work the same way for their level.
* **The length of the conversation.** Each new question carries the earlier messages in the thread along with it. In a very long thread, the cost of each answer grows. A lighter model may also not handle that much context, so WorkLLM may use a stronger one.
* **Documents and memory.** A large file attached to a chat can be sent to the model in full. This can use many credits. A file saved to memory is processed once. After that, only the relevant parts are sent with each question.
* **Reasoning level.** A higher reasoning level uses more credits. Use a lower level for simple tasks.
* **The type of work.** Deep research, web search, and image generation use more credits than a normal answer. Running one prompt on several models side by side uses credits for each model.

### **Typical cost of a task**

The table shows three types of task. A simple task is something like preparing a short report and sending it to a client. A medium task is something like summarizing several documents or analyzing campaign results. A complex task is something like researching a topic across several sources and drafting a full client proposal.

The cost depends on several things together. These are the model you pick, the length of your question and the answer, the documents you attach, whether memory is used, the length of the conversation, and the reasoning level. They are described in the section above.

| Credits per task | Lighter model | Medium model | Higher-end model |
| :- | :- | :- | :- |
| Simple task: prepare a short report and send it to a client | About 5 to 10 | About 8 to 15 | About 10 to 20 |
| Medium task: summarize several documents or analyze campaign results | About 8 to 15 | About 12 to 25 | About 20 to 35 |
| Complex task: research a topic across several sources and draft a full client proposal | About 10 to 30 | About 20 to 40 | About 30 to 50 |

These are estimates, and your results will vary. You can see the credits used after every task.

**Note:** These are high-level estimates and do not include additional factors such as Memory, attached documents, or the length of the conversation. Please keep in mind that your actual credit usage may vary depending on your prompt and use case. 

> WorkLLM cannot show the exact cost before a task runs. The cost also depends on how much the model writes, and that is only known once the answer is complete. To keep costs low, use Recommended with the Low cost tier.

### **How many questions your credits cover**

There is no single number. It depends on how demanding each question is and which model answers it. A short question to a lighter model costs very little. A long question with documents sent to a higher-end model costs much more.

The table shows the credits per question. The number in brackets is how many questions 1,000 credits cover.

<table>
  <colgroup>
    <col width="216" />

    <col width="171" />

    <col width="170" />

    <col width="160" />
  </colgroup>

  <thead>
    <tr>
      <th>Question type</th>
      <th>Lighter model</th>
      <th>Medium model</th>
      <th>Higher-end model</th>
    </tr>
  </thead>

  <tbody>
    <tr>
      <td>Low-end task: a short question with a short answer</td>
      <td>1 to 2 credits (500 to 1,000 questions)</td>
      <td>2 to 8 credits (125 to 500 questions)</td>
      <td>5 to 15 credits (65 to 200 questions)</td>
    </tr>

    <tr>
      <td>Medium task: a one-page answer</td>
      <td>1 to 5 credits (200 to 1,000 questions)</td>
      <td>5 to 20 credits (50 to 200 questions)</td>
      <td>15 to 35 credits (30 to 65 questions)</td>
    </tr>

    <tr>
      <td>High-end task: a long answer with documents, search, or deep reasoning</td>
      <td>5 to 15 credits (65 to 200 questions)</td>
      <td>15 to 40 credits (25 to 65 questions)</td>
      <td>30 to 60 credits (15 to 35 questions)</td>
    </tr>
  </tbody>
</table>

### **Examples for an agency and a consultant**

These examples are estimates for illustration. Your results will depend on your models and your work.

#### **Example 1: a marketing agency**

A marketing agency has 5 people and 10 clients. The team shares 8,000 credits per month.

| Work | Volume per month | Credits each | Credits per month |
| :- | :- | :- | :- |
| Weekly client reports | 40 | 5 to 10 | 200 to 400 |
| Client proposals | 10 | 30 to 50 | 300 to 500 |
| Campaign research | 40 | 10 to 20 | 400 to 800 |
| Competitor research | 30 | 10 to 20 | 300 to 600 |
| Social media content | 200 | 3 to 6 | 600 to 1,200 |
| Campaign copy and variations | 150 | 3 to 6 | 450 to 900 |
| Client emails and communication | 100 | 3 to 6 | 300 to 600 |
| Content briefs | 40 | 5 to 10 | 200 to 400 |
| Document and campaign analysis | 20 | 10 to 20 | 200 to 400 |
| Web searches | 150 | 2 | 300 |
| Image generation | 20 | 10 to 20 | 200 to 400 |
| Agent usage | 20 | 10 to 20 | 200 to 400 |
| **Total** | | | **3,650 to 6,900** |

The agency uses a small part of its pool. It has room for more reports, content, and client work.

#### **Example 2: a solo consultant**

A consultant works alone on a plan with 2,000 credits per month.

| Work | Volume per month | Credits each | Credits per month |
| :- | :- | :- | :- |
| Research briefs | 20 | 10 to 15 | 200 to 300 |
| Client proposals | 10 | 30 to 40 | 300 to 400 |
| Client deliverable drafts | 15 | 15 to 25 | 225 to 375 |
| Document analysis | 20 | 10 to 20 | 200 to 400 |
| Competitor research | 15 | 10 to 20 | 150 to 300 |
| Client emails and communication | 150 | 3 to 5 | 450 to 750 |
| Web searches | 250 | 2 | 500 |
| Meeting and project notes | 45 | 5 to 8 | 225 to 360 |
| Everyday questions | 300 | 3 to 4 | 900 to 1,200 |
| Image generation | 25 | 10 to 15 | 250 to 375 |
| Agent usage | 40 | 10 to 15 | 400 to 600 |
| Client strategy and planning | 10 | 20 to 30 | 200 to 300 |
| **Total** | | | **4,000 to 5,860** |

#### **What can go wrong**

The same consultant attaches a 200-page client report to a long chat. Then they ask a series of short follow-up questions. Each answer can cost much more, because the whole file may be sent with every question.

The fix is simple. Save the report to project memory. WorkLLM then processes it once. After that, only the parts that match each question are sent to the model.

## **Managing your credits**

### **Where to see your credit usage**

You can check your credit usage in three ways.

* **Quick view.** Your workspace credit usage is shown at the top of the screen. You can see it without leaving your workspace.
* **After each answer.** Click the dollar icon under an answer to see which model answered and how many credits that answer used.
* **Detailed view.** Click your profile icon in the bottom-left corner. Then select Credit Usage report. It shows your usage, your remaining credits, and your usage over time. It also shows credits used by model, model provider, task, and agent.

### **Making your credits go further**

These habits help most.

* **Use Recommended.** Leave the model on Recommended. WorkLLM picks the best model for each task based on cost and quality.
* **Choose a cost level.** If cost matters most, pick Recommended Low. It only uses lower-cost models, even for a long thread.
* **Lower the reasoning level.** Use a lower reasoning level for simple tasks. Use a higher level only when the task needs it.
* **Save documents to memory.** If you work with the same files often, save them to your project or folder memory. WorkLLM does not need to process them from scratch each time.
* **Start a new thread for a new topic.** Long threads carry earlier messages along, which raises the cost of each answer.
* **Check your usage.** After each answer, look at the credits used. This shows you what different tasks cost.

### **Which credits are used first, and rollover**

> **For subscription plans only:** Monthly credits roll over for one more month. They expire after that. So each batch of monthly credits is valid for 60 days.

For example, you receive 2,000 credits in January. Credits you do not use in January or February expire at the end of February.

WorkLLM uses your credits in this order.

1. Rolled-over credits first.
2. Monthly credits next.
3. Purchased credits last.

Purchased credits do not expire while your monthly plan is active.

### **Buying more credits, and running out**

You can buy additional credits at 1 cent each. The minimum is 10 dollars for 1,000 credits. The maximum is 1,000 dollars for 100,000 credits. Purchased credits are used after your rolled-over and monthly credits.

If you often need more, a higher plan gives better value. The price per credit is about 1.00 cent on Tier 1. It falls to about 0.71 cent on Tier 5.

When your workspace has used all its available credits, it cannot keep using the platform until more are available. You can buy additional credits or upgrade your plan. The credit meter at the top right shows how many credits your workspace has left. Check it before you start a large task, such as deep research or a long document.

### **Sharing credits across your team**

Credits are shared across the workspace. Everyone in the workspace uses the same credit pool.

> A Workspace Admin can set a credit limit for an individual user. Go to Workspace Settings, then Members. This helps an agency keep one person's usage from using up the whole pool.

## **If your credits dropped more than expected**

Check these common causes first.

* **A premium model.** Recommended may have picked a stronger model for a complex request.
* **A long thread.** Every new question carries the earlier messages along.
* **A large file attached to a chat.** The whole file can be sent with each question.
* **A high reasoning level.** Higher levels use more credits.
* **Deep research, web search, or image generation.** These use more credits than a normal answer.
* **Several models at once.** Each model in a side-by-side run uses its own credits.

To find the cause, open the Credit Usage report and look at usage by model and by task. Click the dollar icon under an answer to see what that answer cost.

To bring costs down, switch to Recommended Low. Start a new thread for a new topic. Move large files into project memory.

If a request failed and you were still charged credits, email [hello@workllm.io](mailto:hello@workllm.io). We will look into it and refund the credits.

## **Refunds and getting help**

You can request a full refund within 7 days of your purchase.

If you have a question about your credit usage, or you want help choosing the right plan for your team, email us at [<u>hello@workllm.io</u>](mailto:hello@workllm.io).

<br />


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.