Using FishXCode with QClaw

First time using FishXCode? Complete Account Setup first, then get your API Key from the FishXCode console.

1. Confirm the Integration Parameters

When QClaw connects to a third-party LLM API, use the OpenAI Compatible protocol first. Even when calling Claude models, fill in the OpenAI compatible endpoint parameters.

Configuration Item	Value
Protocol	OpenAI Compatible
Base URL	https://api.fishxcode.com/v1
API Key	API Key created in the FishXCode console
OpenAI Model	gpt-5.5
Claude Model	claude-opus-4-7

If QClaw requires a full Endpoint, use:

https://api.fishxcode.com/v1/chat/completions

2. Configure Through the Settings UI

If your QClaw version provides a graphical interface, fill it in as follows:

Provider: OpenAI Compatible
Base URL: https://api.fishxcode.com/v1
API Key: sk-你的FishXCode-Token
Model: claude-opus-4-7

To use a GPT model, change Model to:

gpt-5.5

3. Configure Through config.yaml

If your QClaw version uses ~/.qclaw/config.yaml, use the openai-compatible config.yaml scheme below:

llm:
  provider: openai-compatible
  base_url: https://api.fishxcode.com/v1
  api_key: sk-你的FishXCode-Token
  model: claude-opus-4-7
  max_tokens: 8192

When switching to an OpenAI model, only replace model:

llm:
  provider: openai-compatible
  base_url: https://api.fishxcode.com/v1
  api_key: sk-你的FishXCode-Token
  model: gpt-5.5
  max_tokens: 8192

If the client requires a full endpoint URL, change it to:

llm:
  provider: openai-compatible
  url: https://api.fishxcode.com/v1/chat/completions
  api_key: sk-你的FishXCode-Token
  model: claude-opus-4-7
  max_tokens: 8192

Important

Different QClaw versions may use different field names, such as base_url, baseURL, apiBase, or url. If the configuration does not take effect after saving, check the field names and the configuration file path that QClaw actually reads.

4. Configure Through Environment Variables

If QClaw supports environment variables, you can also use:

Linux/macOS

export OPENAI_BASE_URL="https://api.fishxcode.com/v1"
export OPENAI_API_KEY="sk-你的FishXCode-Token"
export QCLAW_MODEL="claude-opus-4-7"
qclaw

Windows PowerShell

$env:OPENAI_BASE_URL="https://api.fishxcode.com/v1"
$env:OPENAI_API_KEY="sk-你的FishXCode-Token"
$env:QCLAW_MODEL="claude-opus-4-7"
qclaw

When you need to use a GPT model, change QCLAW_MODEL to gpt-5.5.

5. Verify the Configuration

First verify that the API Key works with the OpenAI compatible endpoint:

Claude 模型

curl https://api.fishxcode.com/v1/chat/completions \
  -H "content-type: application/json" \
  -H "authorization: Bearer sk-你的FishXCode-Token" \
  -d '{
    "model": "claude-opus-4-7",
    "messages": [
      { "role": "user", "content": "用一句话说明你是谁" }
    ]
  }'

GPT 模型

curl https://api.fishxcode.com/v1/chat/completions \
  -H "content-type: application/json" \
  -H "authorization: Bearer sk-你的FishXCode-Token" \
  -d '{
    "model": "gpt-5.5",
    "messages": [
      { "role": "user", "content": "用一句话说明你是谁" }
    ]
  }'

If curl does not work, QClaw is unlikely to work either. Check the API Key, model permissions, and Base URL first.

1. Save the settings or config.yaml
2. Restart QClaw
3. Select gpt-5.5 or claude-opus-4-7
4. Send a short question, such as 用一句话说明你是谁
5. A normal response means the integration is working

FAQ

404 or endpoint address errors

If filling in a Base URL, use https://api.fishxcode.com/v1. If filling in a full URL, use https://api.fishxcode.com/v1/chat/completions.

Can Claude models be used?

Yes. The prerequisites are that FishXCode currently supports the model and QClaw calls it through the OpenAI Compatible protocol. Use claude-opus-4-7 as the model name.

Authentication fails

Copy the API Key again, make sure there are no extra spaces, and check that the key has not been deleted or disabled.