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

# WorkBuddy

> Add TokenLab as a custom model provider in WorkBuddy

## Overview

<Note>
  **Type**: Desktop coding and task agent

  **Primary Path**: OpenAI-compatible Chat Completions

  **Support Confidence**: Official custom-model path
</Note>

[WorkBuddy](https://www.workbuddy.ai/docs/workbuddy/From-Beginner-to-Expert-Guide/Function-Description/Model) provides a visual custom-model editor for OpenAI-compatible APIs. You do not need to edit its local model file manually.

## Configure TokenLab

1. Open **Settings → Model**.
2. Add a model and select **Custom** as the provider.
3. Enter the following values:

| Field           | Value                        |
| --------------- | ---------------------------- |
| Endpoint URL    | `https://api.tokenlab.sh/v1` |
| API Key         | Your TokenLab API key        |
| Model name      | `kimi-k3`                    |
| Custom Protocol | **Off**                      |

<Warning>
  Keep **Custom Protocol** off. In standard mode WorkBuddy validates the base URL and appends `/chat/completions`. If you turn it on while entering only `https://api.tokenlab.sh/v1`, requests are sent to the wrong path.
</Warning>

## Advanced model settings

For the `kimi-k3` example, use TokenLab's current public capability metadata:

| Setting                  | Value     |
| ------------------------ | --------- |
| Tool calling             | Enabled   |
| Image input              | Enabled   |
| Reasoning mode           | Enabled   |
| Reasoning only           | Disabled  |
| Default reasoning effort | Auto      |
| Input limit              | `1048576` |
| Output limit             | `131072`  |

Click **Save**, then select `kimi-k3` from WorkBuddy's model picker.

## Verify

Start with a small task that exercises both text generation and a safe tool read:

```text theme={null}
Inspect this folder, list its top-level files, and summarize what the project does. Do not modify files.
```

## Troubleshooting

<AccordionGroup>
  <Accordion title="404 or route not found">
    Confirm the endpoint is `https://api.tokenlab.sh/v1` and **Custom Protocol** is off. Do not paste `/chat/completions` when standard mode is selected.
  </Accordion>

  <Accordion title="Model not found">
    Use `kimi-k3`, not `kimi/kimi-k3`. TokenLab exposes model IDs without a provider prefix.
  </Accordion>

  <Accordion title="Agent replies but cannot use tools">
    Enable **Tool calling** and make sure the selected TokenLab model advertises tool-use support.
  </Accordion>
</AccordionGroup>

<Warning>
  Screenshots of the model dialog can reveal endpoints or API keys. Mask credentials before sharing them and rotate any exposed key.
</Warning>