free-claude-code-proxy

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

free-claude-code-proxy

Skill by ara.so — Claude Code Skills collection.

Free Claude Code is a local proxy server that intercepts Anthropic Messages API calls from Claude Code and routes them to alternative providers like NVIDIA NIM, Kimi, Wafer, OpenRouter, DeepSeek, LM Studio, llama.cpp, Ollama, OpenCode Zen, or Z.ai. It maintains Claude Code's client-side protocol while letting you choose free, paid, or local models with full streaming, tool use, and reasoning block support.

由ara.so开发的Skill — Claude Code Skills合集。

Free Claude Code是一个本地代理服务器，它拦截来自Claude Code的Anthropic Messages API请求，并将其路由至其他提供商，如NVIDIA NIM、Kimi、Wafer、OpenRouter、DeepSeek、LM Studio、llama.cpp、Ollama、OpenCode Zen或Z.ai。它保留了Claude Code的客户端协议，同时允许你选择免费、付费或本地模型，全面支持流式传输、工具调用和推理块功能。

Installation

安装

Prerequisites

前提条件

Install uv and Python 3.14:

bash

undefined

安装uv和Python 3.14：

bash

undefined

macOS/Linux

curl -LsSf https://astral.sh/uv/install.sh | sh uv self update uv python install 3.14

Windows PowerShell

powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex" uv self update uv python install 3.14

undefined

powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex" uv self update uv python install 3.14

undefined

Install Claude Code

安装Claude Code

bash

npm install -g @anthropic-ai/claude-code

bash

npm install -g @anthropic-ai/claude-code

Install Free Claude Code Proxy

安装Free Claude Code Proxy

bash

uv tool install --force git+https://github.com/Alishahryar1/free-claude-code.git

Update to latest version with the same command.

bash

uv tool install --force git+https://github.com/Alishahryar1/free-claude-code.git

使用相同命令更新至最新版本。

Starting the Proxy

启动代理

bash

fcc-server

Output shows:

Server URL: http://127.0.0.1:8082
Admin UI: http://127.0.0.1:8082/admin

The server runs on port

by default (configurable via

PORT

env var).

bash

fcc-server

输出示例：

Server URL: http://127.0.0.1:8082
Admin UI: http://127.0.0.1:8082/admin

服务器默认运行在端口

（可通过

PORT

环境变量配置）。

Admin UI Configuration

Admin UI配置

Open

http://127.0.0.1:8082/admin

in your browser to configure providers and models through a local web interface.

在浏览器中打开

http://127.0.0.1:8082/admin

，通过本地Web界面配置提供商和模型。

Quick Setup: NVIDIA NIM

快速设置：NVIDIA NIM

Get API key from https://build.nvidia.com/settings/api-keys
In Admin UI, paste into
```
NVIDIA_NIM_API_KEY
```
Click Validate then Apply
Default
```
MODEL
```
is already set to
```
nvidia_nim/z-ai/glm4.7
```

从https://build.nvidia.com/settings/api-keys获取API密钥
在Admin UI中，将密钥粘贴至
```
NVIDIA_NIM_API_KEY
```
字段
点击Validate（验证）然后Apply（应用）
默认
```
MODEL
```
已设置为
```
nvidia_nim/z-ai/glm4.7
```

Environment Variables

环境变量

The Admin UI manages these settings (or set via

.env

file):

bash

undefined

Admin UI会管理这些设置（也可通过

.env

文件配置）：

bash

undefined

Server config

服务器配置

PORT=8082 AUTH_TOKEN=freecc LOG_LEVEL=INFO

Provider API keys (set one or more)

提供商API密钥（设置一个或多个）

NVIDIA_NIM_API_KEY=nvapi-xxx KIMI_API_KEY=sk-xxx WAFER_API_KEY=sk-xxx OPENROUTER_API_KEY=sk-xxx DEEPSEEK_API_KEY=sk-xxx OPENCODE_API_KEY=sk-xxx ZAI_API_KEY=sk-xxx

Local provider URLs

本地提供商URL

LM_STUDIO_BASE_URL=http://localhost:1234 LLAMACPP_BASE_URL=http://localhost:8080 OLLAMA_BASE_URL=http://localhost:11434

Model routing (prefix with provider/)

模型路由（以提供商名称为前缀）

MODEL=nvidia_nim/z-ai/glm4.7 MODEL_OPUS=nvidia_nim/moonshotai/kimi-k2.5 MODEL_SONNET=open_router/deepseek/deepseek-r1-0528:free MODEL_HAIKU=lmstudio/unsloth/GLM-4.7-Flash-GGUF

undefined

MODEL=nvidia_nim/z-ai/glm4.7 MODEL_OPUS=nvidia_nim/moonshotai/kimi-k2.5 MODEL_SONNET=open_router/deepseek/deepseek-r1-0528:free MODEL_HAIKU=lmstudio/unsloth/GLM-4.7-Flash-GGUF

undefined

Provider Configuration

提供商配置

NVIDIA NIM

bash

undefined

bash

undefined

In Admin UI or .env

在Admin UI或.env中配置

NVIDIA_NIM_API_KEY=nvapi-your-key-here MODEL=nvidia_nim/z-ai/glm4.7


Popular models:
- `nvidia_nim/z-ai/glm4.7`
- `nvidia_nim/z-ai/glm5`
- `nvidia_nim/moonshotai/kimi-k2.5`
- `nvidia_nim/minimaxai/minimax-m2.5`

NVIDIA_NIM_API_KEY=nvapi-your-key-here MODEL=nvidia_nim/z-ai/glm4.7


热门模型：
- `nvidia_nim/z-ai/glm4.7`
- `nvidia_nim/z-ai/glm5`
- `nvidia_nim/moonshotai/kimi-k2.5`
- `nvidia_nim/minimaxai/minimax-m2.5`

Kimi

bash

KIMI_API_KEY=sk-your-key-here
MODEL=kimi/kimi-k2.5

bash

KIMI_API_KEY=sk-your-key-here
MODEL=kimi/kimi-k2.5

Wafer

bash

WAFER_API_KEY=sk-your-key-here
MODEL=wafer/DeepSeek-V4-Pro

Popular models:

```
wafer/DeepSeek-V4-Pro
```
```
wafer/MiniMax-M2.7
```
```
wafer/Qwen3.5-397B-A17B
```
```
wafer/GLM-5.1
```

bash

WAFER_API_KEY=sk-your-key-here
MODEL=wafer/DeepSeek-V4-Pro

热门模型：

```
wafer/DeepSeek-V4-Pro
```
```
wafer/MiniMax-M2.7
```
```
wafer/Qwen3.5-397B-A17B
```
```
wafer/GLM-5.1
```

OpenRouter

bash

OPENROUTER_API_KEY=sk-your-key-here
MODEL=open_router/stepfun/step-3.5-flash:free

Browse models at https://openrouter.ai/models

bash

OPENROUTER_API_KEY=sk-your-key-here
MODEL=open_router/stepfun/step-3.5-flash:free

访问https://openrouter.ai/models浏览更多模型

DeepSeek

bash

DEEPSEEK_API_KEY=sk-your-key-here
MODEL=deepseek/deepseek-chat

Uses Anthropic-compatible endpoint, not OpenAI chat completions.

bash

DEEPSEEK_API_KEY=sk-your-key-here
MODEL=deepseek/deepseek-chat

使用与Anthropic兼容的端点，而非OpenAI聊天补全端点。

LM Studio

bash

LM_STUDIO_BASE_URL=http://localhost:1234
MODEL=lmstudio/your-model-name

Start LM Studio server and load a model with tool-use support.

bash

LM_STUDIO_BASE_URL=http://localhost:1234
MODEL=lmstudio/your-model-name

启动LM Studio服务器并加载支持工具调用的模型。

llama.cpp

bash

LLAMACPP_BASE_URL=http://localhost:8080
MODEL=llamacpp/your-model

Start

llama-server

with sufficient context:

bash

llama-server --model model.gguf --ctx-size 32768 --port 8080

bash

LLAMACPP_BASE_URL=http://localhost:8080
MODEL=llamacpp/your-model

启动

llama-server

并设置足够的上下文长度：

bash

llama-server --model model.gguf --ctx-size 32768 --port 8080

Ollama

bash

OLLAMA_BASE_URL=http://localhost:11434
MODEL=ollama/llama3.1

Pull and serve model:

bash

ollama pull llama3.1
ollama serve

bash

OLLAMA_BASE_URL=http://localhost:11434
MODEL=ollama/llama3.1

拉取并运行模型：

bash

ollama pull llama3.1
ollama serve

OpenCode Zen

bash

OPENCODE_API_KEY=sk-your-key-here
MODEL=opencode/gpt-5.3-codex

Popular models:

```
opencode/gpt-5.3-codex
```
```
opencode/claude-sonnet-4
```
```
opencode/deepseek-v4-flash-free
```
(free)
```
opencode/big-pickle
```
(free)

bash

OPENCODE_API_KEY=sk-your-key-here
MODEL=opencode/gpt-5.3-codex

热门模型：

```
opencode/gpt-5.3-codex
```
```
opencode/claude-sonnet-4
```
```
opencode/deepseek-v4-flash-free
```
（免费）
```
opencode/big-pickle
```
（免费）

Z.ai

bash

ZAI_API_KEY=sk-your-key-here
MODEL=zai/glm-5.1

Models:

```
zai/glm-5.1
```
```
zai/glm-5-turbo
```

bash

ZAI_API_KEY=sk-your-key-here
MODEL=zai/glm-5.1

模型列表：

```
zai/glm-5.1
```
```
zai/glm-5-turbo
```

Connecting Claude Code Clients

连接Claude Code客户端

CLI (Recommended)

CLI（推荐方式）

bash

fcc-claude

This launcher:

Reads current port and auth token from Admin UI config
Sets
```
ANTHROPIC_BASE_URL
```
and
```
ANTHROPIC_AUTH_TOKEN
```
Launches the real
```
claude
```
command

Keep

fcc-server

running in another terminal.

bash

fcc-claude

该启动器会：

从Admin UI配置中读取当前端口和认证令牌
设置
```
ANTHROPIC_BASE_URL
```
和
```
ANTHROPIC_AUTH_TOKEN
```
环境变量
启动原生
```
claude
```
命令

请在另一个终端保持

fcc-server

运行。

Manual CLI Setup

手动CLI设置

bash

export ANTHROPIC_BASE_URL=http://localhost:8082
export ANTHROPIC_AUTH_TOKEN=freecc
export CLAUDE_CODE_ENABLE_GATEWAY_MODEL_DISCOVERY=1
claude

bash

export ANTHROPIC_BASE_URL=http://localhost:8082
export ANTHROPIC_AUTH_TOKEN=freecc
export CLAUDE_CODE_ENABLE_GATEWAY_MODEL_DISCOVERY=1
claude

VS Code Extension

VS Code扩展

Open Settings → search

claude-code.environmentVariables

→ Edit in

settings.json

json

{
  "claudeCode.environmentVariables": [
    { "name": "ANTHROPIC_BASE_URL", "value": "http://localhost:8082" },
    { "name": "ANTHROPIC_AUTH_TOKEN", "value": "freecc" },
    { "name": "CLAUDE_CODE_ENABLE_GATEWAY_MODEL_DISCOVERY", "value": "1" }
  ]
}

Reload VS Code window.

打开设置 → 搜索

claude-code.environmentVariables

→ 在

settings.json

中编辑：

json

{
  "claudeCode.environmentVariables": [
    { "name": "ANTHROPIC_BASE_URL", "value": "http://localhost:8082" },
    { "name": "ANTHROPIC_AUTH_TOKEN", "value": "freecc" },
    { "name": "CLAUDE_CODE_ENABLE_GATEWAY_MODEL_DISCOVERY", "value": "1" }
  ]
}

重新加载VS Code窗口。

JetBrains ACP

Edit installed ACP config:

Windows:

C:\Users\%USERNAME%\AppData\Roaming\JetBrains\acp-agents\installed.json

Linux/macOS:
```
~/.jetbrains/acp.json
```

Add environment to

acp.registry.claude-acp

json

{
  "env": {
    "ANTHROPIC_BASE_URL": "http://localhost:8082",
    "ANTHROPIC_AUTH_TOKEN": "freecc",
    "CLAUDE_CODE_ENABLE_GATEWAY_MODEL_DISCOVERY": "1"
  }
}

Restart IDE.

编辑已安装的ACP配置文件：

Windows：

C:\Users\%USERNAME%\AppData\Roaming\JetBrains\acp-agents\installed.json

Linux/macOS：
```
~/.jetbrains/acp.json
```

在

acp.registry.claude-acp

中添加环境变量：

json

{
  "env": {
    "ANTHROPIC_BASE_URL": "http://localhost:8082",
    "ANTHROPIC_AUTH_TOKEN": "freecc",
    "CLAUDE_CODE_ENABLE_GATEWAY_MODEL_DISCOVERY": "1"
  }
}

重启IDE。

Per-Tier Model Routing

按模型层级路由

Route different Claude Code model tiers to different providers:

bash

undefined

将不同Claude Code模型层级路由至不同提供商：

bash

undefined

Fallback

兜底模型

MODEL=zai/glm-5.1

Override specific tiers

覆盖特定层级

MODEL_OPUS=nvidia_nim/moonshotai/kimi-k2.5 MODEL_SONNET=open_router/deepseek/deepseek-r1-0528:free MODEL_HAIKU=lmstudio/unsloth/GLM-4.7-Flash-GGUF


Leave tier blank to inherit from `MODEL`.

MODEL_OPUS=nvidia_nim/moonshotai/kimi-k2.5 MODEL_SONNET=open_router/deepseek/deepseek-r1-0528:free MODEL_HAIKU=lmstudio/unsloth/GLM-4.7-Flash-GGUF


若层级留空，则继承`MODEL`的设置。

Model Picker Support

模型选择器支持

With

CLAUDE_CODE_ENABLE_GATEWAY_MODEL_DISCOVERY=1

, Claude Code's

/model

picker shows models from the proxy's

/v1/models

endpoint.

The proxy returns configured models based on:

Single provider: lists that provider's models
Multiple providers: shows models from all configured providers
Blank
```
MODEL
```
: returns empty list

Example: If you configure

nvidia_nim

and

open_router

, the picker shows models from both.

设置

CLAUDE_CODE_ENABLE_GATEWAY_MODEL_DISCOVERY=1

后，Claude Code的

/model

选择器会显示代理

/v1/models

端点返回的模型。

代理会根据以下情况返回配置的模型：

单一提供商：列出该提供商的所有模型
多个提供商：显示所有已配置提供商的模型
```
MODEL
```
为空：返回空列表

示例：若你配置了

nvidia_nim

和

open_router

，选择器会显示这两个提供商的所有模型。

Code Examples

代码示例

Python: Direct Proxy Request

Python：直接向代理发送请求

python

import os
import requests

url = "http://localhost:8082/v1/messages"
headers = {
    "x-api-key": os.getenv("AUTH_TOKEN", "freecc"),
    "anthropic-version": "2023-06-01",
    "content-type": "application/json"
}
payload = {
    "model": "claude-3-5-sonnet-20241022",
    "max_tokens": 1024,
    "messages": [
        {"role": "user", "content": "Write a Python hello world"}
    ]
}

response = requests.post(url, json=payload, headers=headers, stream=True)
for line in response.iter_lines():
    if line:
        print(line.decode())

python

import os
import requests

url = "http://localhost:8082/v1/messages"
headers = {
    "x-api-key": os.getenv("AUTH_TOKEN", "freecc"),
    "anthropic-version": "2023-06-01",
    "content-type": "application/json"
}
payload = {
    "model": "claude-3-5-sonnet-20241022",
    "max_tokens": 1024,
    "messages": [
        {"role": "user", "content": "Write a Python hello world"}
    ]
}

response = requests.post(url, json=payload, headers=headers, stream=True)
for line in response.iter_lines():
    if line:
        print(line.decode())

Python: Check Available Models

Python：查看可用模型

python

import requests

response = requests.get(
    "http://localhost:8082/v1/models",
    headers={"x-api-key": "freecc"}
)
models = response.json()
for model in models.get("data", []):
    print(f"{model['id']}: {model['name']}")

python

import requests

response = requests.get(
    "http://localhost:8082/v1/models",
    headers={"x-api-key": "freecc"}
)
models = response.json()
for model in models.get("data", []):
    print(f"{model['id']}: {model['name']}")

Shell: Test Proxy Streaming

Shell：测试代理流式传输

bash

curl -N http://localhost:8082/v1/messages \
  -H "x-api-key: freecc" \
  -H "anthropic-version: 2023-06-01" \
  -H "content-type: application/json" \
  -d '{
    "model": "claude-3-5-sonnet-20241022",
    "max_tokens": 1024,
    "messages": [{"role": "user", "content": "Hello"}],
    "stream": true
  }'

bash

curl -N http://localhost:8082/v1/messages \
  -H "x-api-key: freecc" \
  -H "anthropic-version: 2023-06-01" \
  -H "content-type: application/json" \
  -d '{
    "model": "claude-3-5-sonnet-20241022",
    "max_tokens": 1024,
    "messages": [{"role": "user", "content": "Hello"}],
    "stream": true
  }'

JavaScript: Use Proxy in Custom Client

JavaScript：在自定义客户端中使用代理

javascript

const response = await fetch('http://localhost:8082/v1/messages', {
  method: 'POST',
  headers: {
    'x-api-key': process.env.AUTH_TOKEN || 'freecc',
    'anthropic-version': '2023-06-01',
    'content-type': 'application/json'
  },
  body: JSON.stringify({
    model: 'claude-3-5-sonnet-20241022',
    max_tokens: 1024,
    messages: [{role: 'user', content: 'Explain async/await'}]
  })
});

const data = await response.json();
console.log(data.content[0].text);

javascript

const response = await fetch('http://localhost:8082/v1/messages', {
  method: 'POST',
  headers: {
    'x-api-key': process.env.AUTH_TOKEN || 'freecc',
    'anthropic-version': '2023-06-01',
    'content-type': 'application/json'
  },
  body: JSON.stringify({
    model: 'claude-3-5-sonnet-20241022',
    max_tokens: 1024,
    messages: [{role: 'user', content: 'Explain async/await'}]
  })
});

const data = await response.json();
console.log(data.content[0].text);

Common Workflows

常见工作流

Switch Between Providers

切换提供商

Open Admin UI:
```
http://localhost:8082/admin
```
Update API key for new provider
Change
```
MODEL
```
to new provider prefix (e.g.,
```
kimi/kimi-k2.5
```
)
Click Validate then Apply
Proxy restarts automatically

No need to restart

fcc-claude

— it reads config on each launch.

打开Admin UI：
```
http://localhost:8082/admin
```
更新新提供商的API密钥
将
```
MODEL
```
改为新提供商的前缀（例如
```
kimi/kimi-k2.5
```
）
点击Validate然后Apply
代理会自动重启

无需重启

fcc-claude

——它会在每次启动时读取最新配置。

Use Local Model with Tool Support

使用支持工具调用的本地模型

bash

undefined

bash

undefined

Start LM Studio with a tool-capable model

启动LM Studio并加载支持工具调用的模型

In Admin UI:

在Admin UI中配置：

LM_STUDIO_BASE_URL=http://localhost:1234 MODEL=lmstudio/deepseek-coder-6.7b-instruct

Or for llama.cpp:

或者使用llama.cpp：

llama-server --model deepseek-coder-6.7b-instruct.Q4_K_M.gguf
--ctx-size 16384 --port 8080

LLAMACPP_BASE_URL=http://localhost:8080 MODEL=llamacpp/deepseek-coder-6.7b-instruct

undefined

llama-server --model deepseek-coder-6.7b-instruct.Q4_K_M.gguf
--ctx-size 16384 --port 8080

LLAMACPP_BASE_URL=http://localhost:8080 MODEL=llamacpp/deepseek-coder-6.7b-instruct

undefined

Mix Free and Paid Models

混合使用免费和付费模型

bash

undefined

bash

undefined

Free tier for Haiku

Haiku层级使用免费模型

MODEL_HAIKU=open_router/deepseek/deepseek-r1-0528:free

Paid NVIDIA NIM for Opus

Opus层级使用付费NVIDIA NIM模型

MODEL_OPUS=nvidia_nim/moonshotai/kimi-k2.5

Fallback to free OpenCode

兜底使用免费OpenCode模型

MODEL=opencode/big-pickle

undefined

MODEL=opencode/big-pickle

undefined

Debug Proxy Requests

调试代理请求

Set log level in Admin UI or

.env

bash

LOG_LEVEL=DEBUG

Restart proxy to see full request/response logs.

在Admin UI或

.env

中设置日志级别：

bash

LOG_LEVEL=DEBUG

重启代理后即可查看完整的请求/响应日志。

Troubleshooting

故障排除

Proxy returns 401 Unauthorized

代理返回401 Unauthorized（未授权）

Check
```
AUTH_TOKEN
```
matches in proxy and client
Verify
```
ANTHROPIC_AUTH_TOKEN
```
env var in Claude Code client
Admin UI shows current
```
AUTH_TOKEN
```
value

检查代理和客户端的
```
AUTH_TOKEN
```
是否一致
验证Claude Code客户端中的
```
ANTHROPIC_AUTH_TOKEN
```
环境变量
Admin UI会显示当前的
```
AUTH_TOKEN
```
值

Claude Code says "Invalid API key"

Claude Code提示"Invalid API key"（无效API密钥）

Ensure provider API key is set correctly in Admin UI
Click Validate to test the key
Check provider dashboard for key status

确保提供商API密钥在Admin UI中设置正确
点击Validate测试密钥有效性
检查提供商控制台中的密钥状态

Local model returns HTTP 400

本地模型返回HTTP 400错误

For llama.cpp/LM Studio:

Increase
```
--ctx-size
```
for llama.cpp
Verify model supports tool use
Check server logs for context length errors
Try a smaller conversation history

对于llama.cpp/LM Studio：

增大llama.cpp的
```
--ctx-size
```
参数
验证模型是否支持工具调用
查看服务器日志中的上下文长度错误
尝试缩短对话历史

Model picker shows no models

模型选择器无模型显示

Verify

CLAUDE_CODE_ENABLE_GATEWAY_MODEL_DISCOVERY=1

is set

Check that at least one provider is configured with valid credentials
Visit
```
http://localhost:8082/v1/models
```
to see raw response

确认已设置

CLAUDE_CODE_ENABLE_GATEWAY_MODEL_DISCOVERY=1

检查至少有一个提供商配置了有效凭证
访问
```
http://localhost:8082/v1/models
```
查看原始响应

VS Code extension doesn't use proxy

VS Code扩展未使用代理

Reload VS Code window after changing
```
settings.json
```
Check Output panel → Claude Code for environment variable logs
Verify
```
claudeCode.environmentVariables
```
syntax is correct JSON array

修改
```
settings.json
```
后重新加载VS Code窗口
查看输出面板 → Claude Code中的环境变量日志
验证
```
claudeCode.environmentVariables
```
的JSON数组语法是否正确

Admin UI changes don't apply

Admin UI更改未生效

Click Apply after Validate
Proxy auto-restarts on runtime setting changes
Check terminal for restart confirmation message

点击Validate后再点击Apply
代理会在运行时设置更改后自动重启
查看终端中的重启确认消息

Port 8082 already in use

端口8082已被占用

Change port in Admin UI or

.env

bash

PORT=8083

Restart proxy and update client

ANTHROPIC_BASE_URL

accordingly.

在Admin UI或

.env

中修改端口：

bash

PORT=8083

重启代理并相应更新客户端的

ANTHROPIC_BASE_URL

。

llama.cpp/Ollama not found

无法连接llama.cpp/Ollama

Verify server is running:

curl http://localhost:8080/health

(llama.cpp) or

curl http://localhost:11434

(Ollama)

Check
```
*_BASE_URL
```
doesn't include
```
/v1
```
suffix for Ollama
Ensure firewall allows localhost connections

验证服务器是否运行：

curl http://localhost:8080/health

（llama.cpp）或

curl http://localhost:11434

（Ollama）

检查Ollama的
```
*_BASE_URL
```
是否包含
```
/v1
```
后缀
确保防火墙允许本地连接

Advanced Features

高级功能

Discord Bot Integration

Discord机器人集成

bash

undefined

bash

undefined

In Admin UI or .env

在Admin UI或.env中配置

MESSAGING_PLATFORM=discord DISCORD_BOT_TOKEN=your-bot-token ALLOWED_DISCORD_CHANNELS=123456789,987654321 CLAUDE_WORKSPACE=./agent_workspace


Start bot:
```bash
fcc-bot

MESSAGING_PLATFORM=discord DISCORD_BOT_TOKEN=your-bot-token ALLOWED_DISCORD_CHANNELS=123456789,987654321 CLAUDE_WORKSPACE=./agent_workspace


启动机器人：
```bash
fcc-bot

Voice Transcription

语音转录

bash

undefined

bash

undefined

Local Whisper

使用本地Whisper

VOICE_PROVIDER=whisper WHISPER_MODEL=base

Or NVIDIA NIM

或者使用NVIDIA NIM

VOICE_PROVIDER=nvidia_nim NVIDIA_NIM_API_KEY=nvapi-xxx

undefined

VOICE_PROVIDER=nvidia_nim NVIDIA_NIM_API_KEY=nvapi-xxx

undefined

Health Check Endpoint

健康检查端点

bash

curl http://localhost:8082/health

Returns proxy status and configured provider.

bash

curl http://localhost:8082/health

返回代理状态和已配置的提供商信息。

Request Optimization

请求优化

The proxy automatically:

Strips unsupported parameters for local models
Converts thinking blocks for providers without native support
Handles streaming SSE format differences
Retries with fallback models on provider errors

代理会自动执行以下优化：

为本地模型移除不支持的参数
为不原生支持推理块的提供商转换推理块格式
处理流式SSE格式差异
提供商出错时自动重试并使用兜底模型

Development

开发

Clone and install for development:

bash

git clone https://github.com/Alishahryar1/free-claude-code.git
cd free-claude-code
uv sync
uv run pytest

Run from source:

bash

uv run fcc-server

The project uses:

```
uv
```
for dependency management
```
pytest
```
for testing
```
ruff
```
for linting/formatting
```
loguru
```
for logging
```
ty
```
for type checking

克隆项目并安装开发依赖：

bash

git clone https://github.com/Alishahryar1/free-claude-code.git
cd free-claude-code
uv sync
uv run pytest

从源码运行：

bash

uv run fcc-server

项目使用的工具：

```
uv
```
：依赖管理
```
pytest
```
：测试框架
```
ruff
```
：代码检查/格式化
```
loguru
```
：日志工具
```
ty
```
：类型检查