WorkBuddy

WorkBuddy is Tencent’s AI office agent for document work, data analysis, and code development through natural language. This guide connects Step Plan to WorkBuddy as a custom model service, so you can call Step models with your Step Plan quota.

Prerequisites

An active Step Plan subscription and an API key from the StepFun console.

Add the model in the app

Open model settings

In WorkBuddy, open Settings, select Models, and click Add Model.

Enter the connection details

This guide uses Step 5 Preview as the example. Other models available with Step Plan can be added the same way; see Step Plan overview for the model list.

Field	Value
Provider	Custom (OpenAI-compatible API)
Base URL	`https://api.stepfun.ai/step_plan/v1`
API Key	The key created in the StepFun console
Model name	`step-5-preview`

WorkBuddy appends /chat/completions to the base URL, so enter the base URL only.

Set model capabilities

Open Advanced settings and configure the following. Leave other options at their defaults.

Setting	Value
Tool calling	On, for file reads and other local tasks
Image input	On
Reasoning	On
Custom protocol	Off
Default reasoning effort	`medium`
Available reasoning effort (if shown)	`low`, `medium`, `high`
Max input tokens (if shown)	`1000000`
Max output tokens (if shown)	`64000`

Start with medium and adjust the effort for each task.

Save and send a first message

Click Test connection, then click Save.
In the chat window, select your model from the custom section of the model picker.
Send this message:

Reply with exactly: connected

A reply containing “connected” confirms the in-app setup.

Add the model through the configuration file

Besides the in-app settings, you can add a model by editing WorkBuddy’s configuration file. Use the path shown next to Local configuration file in WorkBuddy settings; on macOS it is usually ~/.workbuddy-ai/models.json. If the file does not exist, add a model once through the app to create it. When the file already lists models, append the new object to the array.

[
  {
    "id": "step-5-preview",
    "name": "Step 5 Preview",
    "vendor": "Custom",
    "url": "https://api.stepfun.ai/step_plan/v1",
    "apiKey": "YOUR_STEP_API_KEY",
    "supportsToolCall": true,
    "supportsImages": true,
    "supportsReasoning": true,
    "useCustomProtocol": false
  }
]

Enter the base URL in url without /chat/completions. Save the file as UTF-8 JSON. The model appears in the app under the name value. If it does not appear, check the path and the JSON syntax, quit WorkBuddy completely, and open it again.

Verify tool calling

After the first reply, open a local test folder, send the request below, and grant folder access when WorkBuddy prompts.

List the top-level files and folders in the current directory and tell me how many items there are. Only read; do not modify or delete anything.

Confirm that WorkBuddy called the file tool and that the result matches the folder contents.

Verify the API directly

To separate app configuration problems from service problems, run the following command in a macOS terminal. After read -s STEP_API_KEY, paste the key and press Return; the input is not echoed.

read -s STEP_API_KEY
printf '\n'
curl --silent --show-error \
  https://api.stepfun.ai/step_plan/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer ${STEP_API_KEY}" \
  -d '{
    "model": "step-5-preview",
    "messages": [{"role": "user", "content": "Reply with exactly OK."}],
    "reasoning_effort": "low",
    "max_tokens": 1024,
    "stream": false
  }'
unset STEP_API_KEY

choices[0].message.content equal to OK means the endpoint and the key work. If the request fails, read the error field. This request verifies the endpoint and the key; the in-app setup still needs a message sent from WorkBuddy.

Troubleshooting

Symptom	Cause	What to do
Test connection reports an invalid key or a permission error	The message merges several authentication errors	Check that the key is complete and the account has model access, then read the exact server error with the direct API request above.
The model appears in the picker but a message fails	Another entry is selected, or the configuration did not apply	Select your own entry in the model picker and read the full error message.
404 or model not found	The base URL or model name is incorrect	Set the base URL to `https://api.stepfun.ai/step_plan/v1` and the model name to `step-5-preview`; make sure Custom protocol is off.
The model does not appear after editing `models.json`	The file is not the one WorkBuddy reads, the JSON is invalid, or the model is missing from the array	Check the path, validate the JSON, and reopen WorkBuddy.
Chat works but no file tool runs	Tool calling is off, or folder access was not granted	Turn on Tool calling and allow WorkBuddy to access the test folder.
402 or `quota_exceeded`	The base URL points to the pay-as-you-go endpoint, or the account has no quota	Use `https://api.stepfun.ai/step_plan/v1`. Step Plan quota applies to the `step_plan` endpoint; the standard `/v1` endpoint bills against the account balance.

Step Plan quickstart: account setup, endpoints, and tool integration settings.
Reasoning models: reasoning_effort and related reasoning settings.
Step 5 Preview: context window, input types, and model capabilities.

Integrations

Migration

Text & Reasoning

Requests & Context

Agents & Tools

Vision

Audio

Errors & Troubleshooting

Prerequisites

Add the model in the app

Add the model through the configuration file

Verify tool calling

Verify the API directly

Troubleshooting

​Prerequisites

​Add the model in the app

​Add the model through the configuration file

​Verify tool calling

​Verify the API directly

​Troubleshooting

​Related documentation

Prerequisites

Add the model in the app

Add the model through the configuration file

Verify tool calling

Verify the API directly

Troubleshooting

Related documentation