@sekora_ai/claude-toggle 1.0.0 → 1.1.0

package/PROVIDERS.md

@@ -141,6 +141,107 @@ Then use it:
 }
 ```
 
+## CCR-Routed Providers
+
+For providers requiring request/response transformation (like Google Gemini), claude-toggle can use [Claude Code Router (CCR)](https://github.com/musistudio/claude-code-router) as a middleware proxy.
+
+### CCR Provider Schema
+
+CCR providers use additional fields:
+
+```json
+{
+  "version": "2.0",
+  "ccr": {
+    "enabled": true,
+    "host": "127.0.0.1",
+    "port": 3456,
+    "log_level": "info",
+    "api_timeout_ms": 600000,
+    "auto_install": true,
+    "auto_start": true
+  },
+  "providers": {
+    "gemini-ccr": {
+      "description": "Google Gemini 3 via Claude Code Router",
+      "enabled": true,
+      "ccr_provider": true,
+      "ccr_config": {
+        "name": "gemini",
+        "api_base_url": "https://generativelanguage.googleapis.com/v1beta/models/",
+        "api_key_env": "GEMINI_API_KEY",
+        "models": {
+          "default": "gemini-3-pro",
+          "background": "gemini-3-flash",
+          "think": "gemini-3-deep-think"
+        },
+        "transformer": {
+          "use": ["gemini"]
+        }
+      },
+      "routing_modes": {
+        "standard": {
+          "default": "gemini-3-pro",
+          "background": "gemini-3-flash"
+        },
+        "reasoning": {
+          "default": "gemini-3-pro",
+          "think": "gemini-3-deep-think",
+          "background": "gemini-3-flash"
+        },
+        "eco": {
+          "default": "gemini-3-flash",
+          "background": "gemini-3-flash"
+        }
+      },
+      "validation": {
+        "required_env_vars": ["GEMINI_API_KEY"]
+      }
+    }
+  }
+}
+```
+
+### CCR-Specific Fields
+
+| Field | Type | Description |
+|-------|------|-------------|
+| `ccr_provider` | boolean | Indicates this provider routes through CCR |
+| `ccr_config.name` | string | Provider identifier in CCR config |
+| `ccr_config.api_base_url` | string | API endpoint URL |
+| `ccr_config.api_key_env` | string | Environment variable containing API key |
+| `ccr_config.models` | object | Model mapping for router slots |
+| `ccr_config.transformer` | object | CCR transformer configuration |
+| `routing_modes` | object | Named routing configurations |
+
+### CCR Global Settings (`ccr` section)
+
+| Field | Type | Default | Description |
+|-------|------|---------|-------------|
+| `enabled` | boolean | true | Enable CCR integration |
+| `host` | string | "127.0.0.1" | CCR server host |
+| `port` | number | 3456 | CCR server port |
+| `log_level` | string | "info" | CCR log level |
+| `api_timeout_ms` | number | 600000 | API request timeout |
+| `auto_install` | boolean | true | Auto-install CCR via npm if missing |
+| `auto_start` | boolean | true | Auto-start CCR server when needed |
+
+### Routing Modes
+
+CCR providers can define multiple routing modes:
+
+- **standard**: Default configuration for general tasks
+- **reasoning**: Uses the `think` router for complex reasoning tasks
+- **eco**: Uses fastest/cheapest models for cost-sensitive tasks
+
+Usage:
+
+```bash
+./claude-toggle --provider gemini-ccr --ccr-mode reasoning
+```
+
+See [docs/CCR_INTEGRATION.md](docs/CCR_INTEGRATION.md) for comprehensive CCR documentation.
+
 ## Tips
 
 1. **Variable References**: Use `${VAR_NAME}` in `value` fields to reference environment variables. The script will expand these at runtime.
@@ -153,6 +254,96 @@ Then use it:
 
 5. **Validation**: Use `required_env_vars` to ensure credentials are set before launching. The script will show helpful setup instructions if they're missing.
 
+6. **CCR Providers**: For providers requiring request/response transformation, set `ccr_provider: true` and define `ccr_config`.
+
+7. **LiteLLM Providers**: For routing through [LiteLLM](https://docs.litellm.ai/), point `ANTHROPIC_BASE_URL` to your LiteLLM server and use model names like `gemini/gemini-2.0-flash`. LiteLLM returns actual model names in responses.
+
+## LiteLLM Integration
+
+[LiteLLM](https://docs.litellm.ai/) is a unified API gateway for 100+ AI models, with official support mentioned in [Claude Code's LLM Gateway documentation](https://code.claude.com/docs/en/llm-gateway).
+
+### Why Use LiteLLM?
+
+- **Shows actual model names** in Claude Code UI (e.g., "gemini-2.0-flash" instead of "Opus 4.5")
+- **100+ provider support** - Gemini, OpenAI, DeepSeek, Anthropic, and more
+- **Official Anthropic compatibility** - Native Anthropic API format support
+- **Load balancing and fallbacks** - Production-ready features
+- **Single API key** - Manage all providers through one gateway
+
+### LiteLLM Provider Example
+
+```json
+"litellm-gemini": {
+  "description": "Google Gemini 2 via LiteLLM (shows model name)",
+  "enabled": true,
+  "env_vars": {
+    "ANTHROPIC_API_KEY": {
+      "value": "${GEMINI_API_KEY}",
+      "required": true
+    },
+    "ANTHROPIC_BASE_URL": {
+      "value": "${LITELLM_BASE_URL:-http://localhost:4000}"
+    },
+    "ANTHROPIC_DEFAULT_OPUS_MODEL": {
+      "value": "gemini/gemini-2.0-flash"
+    }
+  },
+  "validation": {
+    "required_env_vars": ["GEMINI_API_KEY"]
+  }
+}
+```
+
+### Setting Up LiteLLM
+
+1. **Install LiteLLM:**
+   ```bash
+   pip install 'litellm[proxy]'
+   ```
+
+2. **Start the proxy server:**
+   ```bash
+   # For Gemini (requires --drop_params for Claude Code compatibility)
+   litellm --model gemini/gemini-2.0-flash --port 4000 --drop_params
+
+   # For OpenAI
+   litellm --model openai/gpt-4o --port 4000
+
+   # Background mode
+   litellm --model gemini/gemini-2.0-flash --port 4000 --drop_params &
+   ```
+
+3. **Set provider API key:**
+   ```bash
+   export GEMINI_API_KEY="your-gemini-key"
+   # or
+   export OPENAI_API_KEY="your-openai-key"
+   ```
+
+4. **Verify LiteLLM is running:**
+   ```bash
+   nc -z localhost 4000 && echo "LiteLLM is running" || echo "LiteLLM is not running"
+   ```
+
+### Available LiteLLM Providers
+
+| Provider | Description | Required Env Var |
+|----------|-------------|------------------|
+| `litellm-gemini` | Google Gemini 2 via LiteLLM | `GEMINI_API_KEY` |
+| `litellm-openai` | OpenAI GPT-4 via LiteLLM | `OPENAI_API_KEY` |
+| `litellm-generic` | Custom model via `LITELLM_MODEL` env var | Provider-specific key |
+
+### Model Name Format
+
+LiteLLM uses the format `provider/model`:
+- `gemini/gemini-2.0-flash`
+- `gemini/gemini-1.5-pro`
+- `openai/gpt-4o`
+- `openai/gpt-4o-mini`
+- `deepseek/deepseek-chat`
+
+See [LiteLLM providers documentation](https://docs.litellm.ai/docs/providers) for the full list.
+
 ## Testing Your Configuration
 
 After adding a new provider, validate it:

package/README.md

@@ -9,6 +9,8 @@ A provider-agnostic utility for toggling between different Claude API providers
 - **Pass-through design** - All Claude Code CLI options work transparently
 - **Actionable error messages** - Missing credentials trigger detailed setup instructions
 - **Extensible** - Add new providers by editing a single JSON file
+- **CCR Integration** - Route through [Claude Code Router](https://github.com/musistudio/claude-code-router) for providers like Gemini
+- **Advanced routing modes** - Standard, reasoning (deep think), and eco modes for CCR providers
 
 ## Prerequisites
 
@@ -30,7 +32,7 @@ This works on macOS, Linux, and Windows (via WSL or Git Bash).
 
 ```bash
 # Clone the repository
-git clone https://github.com/seansekora/claude-toggle.git
+git clone https://gitlab.com/sekora-ai/claude-code/claude-toggle.git
 cd claude-toggle
 chmod +x claude-toggle
 
@@ -89,6 +91,16 @@ All unrecognized options are forwarded to the Claude Code CLI:
 | `-h, --help` | Show help message |
 | `--glm` | Legacy alias for `--provider glm` |
 
+### CCR Options (for CCR-routed providers)
+
+| Option | Description |
+|--------|-------------|
+| `--ccr-mode <mode>` | Select routing mode: `standard`, `reasoning`, `eco` |
+| `--ccr-status` | Show CCR server status |
+| `--ccr-restart` | Restart CCR server |
+| `--ccr-install` | Install CCR globally via npm |
+| `--ccr-stop` | Stop CCR server |
+
 ## Configuration
 
 ### Configuration File Location
@@ -168,10 +180,12 @@ See [PROVIDERS.md](PROVIDERS.md) for detailed instructions on adding new provide
 
 ## Built-in Providers
 
-| Provider | Description | Required Env Var |
-|----------|-------------|------------------|
-| `anthropic` | Anthropic's official API (default) | None (uses default auth) |
-| `glm` | Z.AI GLM models via proxy | `ZAI_API_KEY` |
+| Provider | Description | Required Env Var | Routing |
+|----------|-------------|------------------|---------|
+| `anthropic` | Anthropic's official API (default) | None (uses default auth) | Direct |
+| `glm` | Z.AI GLM models via proxy | `ZAI_API_KEY` | Direct |
+| `gemini-ccr` | Google Gemini via CCR | `GEMINI_API_KEY` | CCR |
+| `litellm-gemini` | Google Gemini via LiteLLM | `GEMINI_API_KEY` | LiteLLM |
 
 ## Examples
 
@@ -203,6 +217,43 @@ export ZAI_API_KEY="your-key-here"
 # Output: Using provider: glm
 ```
 
+### Using Google Gemini via CCR
+
+```bash
+# Set your API key
+export GEMINI_API_KEY="your-key-here"
+
+# Launch with standard mode (default)
+./claude-toggle --provider gemini-ccr
+
+# Launch with reasoning mode (uses gemini-3-deep-think)
+./claude-toggle --provider gemini-ccr --ccr-mode reasoning
+
+# Launch with eco mode (uses gemini-3-flash for cost savings)
+./claude-toggle --provider gemini-ccr --ccr-mode eco
+
+# Check CCR status
+./claude-toggle --ccr-status
+```
+
+### Using Google Gemini via LiteLLM (Shows Model Name)
+
+```bash
+# Install and start LiteLLM
+pip install 'litellm[proxy]'
+litellm --model gemini/gemini-2.0-flash --port 4000 --drop_params &
+
+# Set your API key
+export GEMINI_API_KEY="your-key-here"
+
+# Launch (will show "gemini-2.0-flash" in UI)
+./claude-toggle --provider litellm-gemini
+```
+
+**Note**: The `--drop_params` flag is required for Gemini compatibility with Claude Code. It drops unsupported parameters like `reasoning_effort`.
+
+**Note**: LiteLLM displays the actual model name in the Claude Code UI, unlike CCR which shows "Opus 4.5".
+
 ## CI/CD Setup (For Maintainers)
 
 This project uses GitLab CI/CD to automatically publish to npm when version tags are pushed.
@@ -273,6 +324,7 @@ claude-toggle/
 │   ├── wrapper.js             # Node.js wrapper for npm
 │   └── postinstall.js         # Post-install config setup
 ├── docs/
+│   ├── CCR_INTEGRATION.md     # CCR integration guide
 │   └── NPM_PUBLISHING.md      # Publishing guide
 ├── .claude/
 │   └── commands/
@@ -332,12 +384,12 @@ claude-toggle is a bash script and requires a Unix-like environment:
 1. **Windows Subsystem for Linux (WSL)** - Recommended
    ```bash
    # Install WSL, then inside WSL:
-   npm install -g claude-toggle
+   npm install -g @sekora_ai/claude-toggle
    ```
 
 2. **Git Bash** - With Python 3 installed
    ```bash
-   npm install -g claude-toggle
+   npm install -g @sekora_ai/claude-toggle
    ```
 
 ## Troubleshooting

package/claude-toggle

@@ -9,7 +9,7 @@ set -e
 # =============================================
 # Script Constants
 # =============================================
-SCRIPT_VERSION="1.0.0"
+SCRIPT_VERSION="1.1.0"
 SCRIPT_NAME="$(basename "$0")"
 SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd)"
 
@@ -27,6 +27,15 @@ CLAUDE_ARGS=()
 ACTION="launch"  # launch, list, validate, help
 VERBOSE=false
 
+# CCR (Claude Code Router) Settings
+CCR_CONFIG_DIR="$HOME/.claude-code-router"
+CCR_CONFIG_FILE="$CCR_CONFIG_DIR/config.json"
+CCR_PACKAGE="@musistudio/claude-code-router"
+CCR_DEFAULT_PORT=3456
+CCR_DEFAULT_HOST="127.0.0.1"
+CCR_MODE=""           # standard, reasoning, eco
+CCR_ACTION=""         # status, restart, install, stop
+
 # =============================================
 # Usage Information
 # =============================================
@@ -43,6 +52,13 @@ OPTIONS:
   -h, --help               Show this help message
   --validate               Validate provider configuration without launching
 
+CCR OPTIONS (for CCR-routed providers like gemini-ccr):
+  --ccr-mode <mode>        Select routing mode: standard, reasoning, eco
+  --ccr-status             Show CCR server status
+  --ccr-restart            Restart CCR server
+  --ccr-install            Install CCR globally via npm
+  --ccr-stop               Stop CCR server
+
 EXAMPLES:
   # Launch with default provider (anthropic)
   $SCRIPT_NAME
@@ -50,9 +66,21 @@ EXAMPLES:
   # Launch with specific provider
   $SCRIPT_NAME --provider glm
 
+  # Use Gemini via CCR with standard mode
+  $SCRIPT_NAME --provider gemini-ccr
+
+  # Use Gemini with reasoning mode (deep think)
+  $SCRIPT_NAME --provider gemini-ccr --ccr-mode reasoning
+
+  # Use Gemini with eco mode (fast/cheap)
+  $SCRIPT_NAME --provider gemini-ccr --ccr-mode eco
+
   # List available providers
   $SCRIPT_NAME --list-providers
 
+  # Check CCR server status
+  $SCRIPT_NAME --ccr-status
+
   # Pass additional options to claude
   $SCRIPT_NAME --provider glm --dangerously-skip-permissions
 
@@ -221,6 +249,76 @@ To set up your ANTHROPIC_API_KEY:
 3. Verify the key is set:
    echo $ANTHROPIC_API_KEY
 
+EOF
+      ;;
+    GEMINI_API_KEY)
+      cat >&2 << 'EOF'
+
+To set up your GEMINI_API_KEY:
+
+1. Get your API key from Google AI Studio:
+   - Visit: https://aistudio.google.com/apikey
+   - Sign in with your Google account
+   - Click "Create API Key"
+   - Copy your API key
+
+2. Set the environment variable:
+
+   # For current session only (temporary):
+   export GEMINI_API_KEY="your-api-key-here"
+
+   # For permanent setup (recommended):
+   # Add this line to your shell configuration file:
+
+   # For bash (~/.bashrc or ~/.bash_profile):
+   echo 'export GEMINI_API_KEY="your-api-key-here"' >> ~/.bashrc
+   source ~/.bashrc
+
+   # For zsh (~/.zshrc):
+   echo 'export GEMINI_API_KEY="your-api-key-here"' >> ~/.zshrc
+   source ~/.zshrc
+
+   # For fish (~/.config/fish/config.fish):
+   echo 'set -x GEMINI_API_KEY "your-api-key-here"' >> ~/.config/fish/config.fish
+
+3. Verify the key is set:
+   echo $GEMINI_API_KEY
+
+EOF
+      ;;
+    LITELLM*)
+      cat >&2 << 'EOF'
+
+LiteLLM is not running. To set up LiteLLM:
+
+1. Install LiteLLM:
+   pip install litellm[proxy]
+   # Or: pip3 install litellm[proxy]
+
+2. Start the LiteLLM proxy server:
+   # For Gemini:
+   litellm --model gemini/gemini-2.0-flash --port 4000
+
+   # For OpenAI:
+   litellm --model openai/gpt-4o --port 4000
+
+   # Or run in background:
+   litellm --model gemini/gemini-2.0-flash --port 4000 &
+
+3. Set your provider's API key:
+
+   # For Gemini:
+   export GEMINI_API_KEY="your-gemini-key"
+
+   # For OpenAI:
+   export OPENAI_API_KEY="your-openai-key"
+
+4. Verify LiteLLM is running:
+   curl http://localhost:4000/status
+
+5. Optionally set a custom LiteLLM URL:
+   export LITELLM_BASE_URL="http://localhost:4000"
+
 EOF
       ;;
     *)
@@ -255,6 +353,235 @@ EOF
   esac
 }
 
+# =============================================
+# CCR (Claude Code Router) Management
+# =============================================
+
+# Check if CCR is installed
+check_ccr_installed() {
+  if command -v ccr &> /dev/null; then
+    return 0
+  else
+    return 1
+  fi
+}
+
+# Install CCR via npm
+install_ccr() {
+  echo "Installing Claude Code Router..." >&2
+  if command -v npm &> /dev/null; then
+    if npm install -g "$CCR_PACKAGE" 2>&1 | while read -r line; do
+      echo "  $line" >&2
+    done; then
+      echo "CCR installed successfully" >&2
+      return 0
+    else
+      echo "Error: Failed to install CCR" >&2
+      return 1
+    fi
+  else
+    echo "Error: npm not found. Please install Node.js and npm first." >&2
+    return 1
+  fi
+}
+
+# Check if CCR server is running
+check_ccr_running() {
+  local status_output
+  status_output=$(ccr status 2>&1)
+  if echo "$status_output" | grep -qi "running\|active\|listening\|started"; then
+    return 0
+  else
+    return 1
+  fi
+}
+
+# Start CCR server
+start_ccr() {
+  if check_ccr_running; then
+    [[ "$VERBOSE" == true ]] && echo "CCR already running" >&2
+    return 0
+  fi
+
+  echo "Starting Claude Code Router..." >&2
+  ccr start 2>&1 | while read -r line; do
+    [[ "$VERBOSE" == true ]] && echo "  $line" >&2
+  done
+
+  # Wait for CCR to become available
+  local max_wait=10
+  local waited=0
+  while [[ $waited -lt $max_wait ]]; do
+    if check_ccr_running; then
+      echo "CCR started successfully" >&2
+      return 0
+    fi
+    sleep 1
+    ((waited++))
+  done
+
+  echo "Error: CCR failed to start within ${max_wait}s" >&2
+  return 1
+}
+
+# Stop CCR server
+stop_ccr() {
+  if ! check_ccr_running; then
+    [[ "$VERBOSE" == true ]] && echo "CCR not running" >&2
+    return 0
+  fi
+
+  echo "Stopping Claude Code Router..." >&2
+  ccr stop 2>&1
+  return $?
+}
+
+# Restart CCR server
+restart_ccr() {
+  echo "Restarting Claude Code Router..." >&2
+  ccr restart 2>&1
+  return $?
+}
+
+# Get CCR status
+get_ccr_status() {
+  if ! check_ccr_installed; then
+    echo "CCR Status: Not installed"
+    echo "  Install with: $SCRIPT_NAME --ccr-install"
+    return 1
+  fi
+
+  echo "CCR Status:"
+  ccr status 2>&1 | sed 's/^/  /'
+
+  if [[ -f "$CCR_CONFIG_FILE" ]]; then
+    echo ""
+    echo "Config: $CCR_CONFIG_FILE"
+  else
+    echo ""
+    echo "Config: Not generated (run with CCR provider to generate)"
+  fi
+}
+
+# Generate CCR config.json from providers.json
+generate_ccr_config() {
+  local provider="$1"
+  local routing_mode="${2:-standard}"
+  local config_file="$3"
+
+  # Ensure CCR config directory exists
+  mkdir -p "$CCR_CONFIG_DIR"
+
+  # Generate config using Python
+  python3 << PYTHON_SCRIPT
+import json
+import os
+import sys
+
+config_file = '$config_file'
+provider_name = '$provider'
+routing_mode = '$routing_mode'
+ccr_config_file = '$CCR_CONFIG_FILE'
+
+try:
+    with open(config_file, 'r') as f:
+        data = json.load(f)
+
+    provider = data.get('providers', {}).get(provider_name)
+    if not provider:
+        print(f"Error: Provider '{provider_name}' not found", file=sys.stderr)
+        sys.exit(1)
+
+    if not provider.get('ccr_provider'):
+        print(f"Error: Provider '{provider_name}' is not a CCR provider", file=sys.stderr)
+        sys.exit(1)
+
+    ccr_config = provider.get('ccr_config', {})
+    routing_modes = provider.get('routing_modes', {})
+    global_ccr = data.get('ccr', {})
+
+    # Get routing configuration for selected mode
+    mode_config = routing_modes.get(routing_mode, routing_modes.get('standard', {}))
+
+    # Build CCR config.json
+    ccr_output = {
+        "HOST": global_ccr.get('host', '127.0.0.1'),
+        "PORT": global_ccr.get('port', 3456),
+        "LOG": True,
+        "LOG_LEVEL": global_ccr.get('log_level', 'info'),
+        "API_TIMEOUT_MS": global_ccr.get('api_timeout_ms', 600000),
+        "Providers": [
+            {
+                "name": ccr_config.get('name', provider_name),
+                "api_base_url": ccr_config.get('api_base_url'),
+                "api_key": f"\${{{ccr_config.get('api_key_env')}}}",
+                "models": list(set(
+                    list(ccr_config.get('models', {}).values()) +
+                    list(mode_config.values())
+                )),
+                "transformer": ccr_config.get('transformer', {})
+            }
+        ],
+        "Router": {}
+    }
+
+    # Build Router section from mode_config
+    provider_prefix = ccr_config.get('name', provider_name)
+    for slot, model in mode_config.items():
+        ccr_output["Router"][slot] = f"{provider_prefix},{model}"
+
+    # Ensure at least default is set
+    if 'default' not in ccr_output["Router"]:
+        default_model = ccr_config.get('models', {}).get('default', 'unknown')
+        ccr_output["Router"]["default"] = f"{provider_prefix},{default_model}"
+
+    # Write CCR config
+    with open(ccr_config_file, 'w') as f:
+        json.dump(ccr_output, f, indent=2)
+
+    print(f"Generated CCR config at {ccr_config_file}", file=sys.stderr)
+
+except Exception as e:
+    print(f"Error generating CCR config: {e}", file=sys.stderr)
+    sys.exit(1)
+PYTHON_SCRIPT
+}
+
+# Ensure CCR is ready (install, generate config, start)
+ensure_ccr_ready() {
+  local provider="$1"
+  local routing_mode="$2"
+  local config_file="$3"
+
+  # Check CCR global settings
+  local auto_install
+  auto_install=$(parse_json "$config_file" "ccr.auto_install" 2>/dev/null || echo "true")
+
+  # Install CCR if needed
+  if ! check_ccr_installed; then
+    if [[ "$auto_install" == "true" || "$auto_install" == "True" ]]; then
+      install_ccr || return 1
+    else
+      echo "Error: CCR not installed and auto_install is disabled" >&2
+      echo "Install manually: npm install -g $CCR_PACKAGE" >&2
+      return 1
+    fi
+  fi
+
+  # Generate CCR config
+  generate_ccr_config "$provider" "$routing_mode" "$config_file" || return 1
+
+  # Start or restart CCR
+  if check_ccr_running; then
+    # Config changed, restart
+    restart_ccr || return 1
+  else
+    start_ccr || return 1
+  fi
+
+  return 0
+}
+
 # =============================================
 # Provider Validation
 # =============================================
@@ -299,6 +626,39 @@ validate_provider() {
     fi
   fi
 
+  # Check if LiteLLM server is running for litellm providers
+  if [[ "$provider" == litellm-* ]]; then
+    local litellm_url
+    litellm_url=$(parse_json "$config_file" "providers.$provider.env_vars.ANTHROPIC_BASE_URL.value" 2>/dev/null || echo "http://localhost:4000")
+
+    # Expand ${VAR} references
+    if [[ "$litellm_url" == *\$* ]]; then
+      while [[ "$litellm_url" == *\$* ]]; do
+        local var_name="${litellm_url#*\$\{}"
+        var_name="${var_name%\}*}"
+        var_name="${var_name%%\}*}"
+        local var_value="${!var_name:-}"
+        litellm_url="${litellm_url/\$\{$var_name\*/$var_value}"
+        litellm_url="${litellm_url/\$\{$var_name\}$var_value}"
+      done
+    fi
+
+    # Extract host and port for TCP check
+    local check_url="${litellm_url#*://}"
+    local host="${check_url%:*}"
+    local port="${check_url#*:}"
+
+    # Check if LiteLLM is running using TCP connection (not HTTP endpoint)
+    if ! nc -z -w5 "$host" "$port" 2>/dev/null; then
+      echo "Error: LiteLLM server is not running at $litellm_url" >&2
+      echo "" >&2
+      echo "Tip: LiteLLM should show 'Uvicorn running on http://0.0.0.0:4000' when started" >&2
+      echo "" >&2
+      show_setup_instructions "LITELLM"
+      return 1
+    fi
+  fi
+
   return 0
 }
 
@@ -418,6 +778,27 @@ main() {
         PROVIDER_NAME="glm"
         shift
         ;;
+      # CCR options
+      --ccr-mode)
+        CCR_MODE="$2"
+        shift 2
+        ;;
+      --ccr-status)
+        CCR_ACTION="status"
+        shift
+        ;;
+      --ccr-restart)
+        CCR_ACTION="restart"
+        shift
+        ;;
+      --ccr-install)
+        CCR_ACTION="install"
+        shift
+        ;;
+      --ccr-stop)
+        CCR_ACTION="stop"
+        shift
+        ;;
       # Passthrough to claude
       *)
         CLAUDE_ARGS+=("$1")
@@ -448,6 +829,28 @@ main() {
     exit 0
   fi
 
+  # Handle CCR actions (don't require provider validation)
+  if [[ -n "$CCR_ACTION" ]]; then
+    case "$CCR_ACTION" in
+      status)
+        get_ccr_status
+        exit $?
+        ;;
+      restart)
+        restart_ccr
+        exit $?
+        ;;
+      install)
+        install_ccr
+        exit $?
+        ;;
+      stop)
+        stop_ccr
+        exit $?
+        ;;
+    esac
+  fi
+
   # Validate config file exists for other actions
   if [[ -z "$config_file" ]]; then
     echo "Error: Configuration file not found" >&2
@@ -479,8 +882,32 @@ main() {
     echo "Using provider: $PROVIDER_NAME" >&2
   fi
 
-  # Export the environment variables
-  eval "$(export_provider_vars "$PROVIDER_NAME" "$config_file")"
+  # Check if this is a CCR provider
+  local is_ccr_provider
+  is_ccr_provider=$(parse_json "$config_file" "providers.$PROVIDER_NAME.ccr_provider" 2>/dev/null || echo "false")
+
+  if [[ "$is_ccr_provider" == "true" || "$is_ccr_provider" == "True" ]]; then
+    # Set default routing mode if not specified
+    [[ -z "$CCR_MODE" ]] && CCR_MODE="standard"
+
+    # Ensure CCR is ready (install, config, start)
+    if ! ensure_ccr_ready "$PROVIDER_NAME" "$CCR_MODE" "$config_file"; then
+      exit 1
+    fi
+
+    # Get CCR host/port for environment setup
+    local ccr_host ccr_port
+    ccr_host=$(parse_json "$config_file" "ccr.host" 2>/dev/null || echo "127.0.0.1")
+    ccr_port=$(parse_json "$config_file" "ccr.port" 2>/dev/null || echo "3456")
+
+    # Set environment to route through CCR
+    export ANTHROPIC_BASE_URL="http://${ccr_host}:${ccr_port}"
+
+    [[ "$VERBOSE" == true ]] && echo "Routing through CCR at $ANTHROPIC_BASE_URL (mode: $CCR_MODE)" >&2
+  else
+    # Standard provider - export env vars directly
+    eval "$(export_provider_vars "$PROVIDER_NAME" "$config_file")"
+  fi
 
   # Launch claude with passthrough arguments
   exec claude "${CLAUDE_ARGS[@]}"

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@sekora_ai/claude-toggle",
-  "version": "1.0.0",
+  "version": "1.1.0",
   "description": "Provider-agnostic toggle for Claude Code API providers",
   "keywords": [
     "claude",

package/providers.json

@@ -1,7 +1,16 @@
 {
   "$schema": "http://json-schema.org/draft-07/schema#",
-  "version": "1.0",
+  "version": "2.0",
   "default_provider": "anthropic",
+  "ccr": {
+    "enabled": true,
+    "host": "127.0.0.1",
+    "port": 3456,
+    "log_level": "info",
+    "api_timeout_ms": 600000,
+    "auto_install": true,
+    "auto_start": true
+  },
   "providers": {
     "anthropic": {
       "description": "Anthropic's official API",
@@ -36,6 +45,92 @@
       "validation": {
         "required_env_vars": ["ZAI_API_KEY"]
       }
+    },
+    "litellm-gemini": {
+      "description": "Google Gemini 3 via LiteLLM (shows model name)",
+      "enabled": true,
+      "env_vars": {
+        "ANTHROPIC_API_KEY": {
+          "value": "${GEMINI_API_KEY}",
+          "required": true
+        },
+        "ANTHROPIC_BASE_URL": {
+          "value": "http://localhost:4000"
+        },
+        "ANTHROPIC_DEFAULT_HAIKU_MODEL": {
+          "value": "gemini/gemini-3.0-pro"
+        },
+        "ANTHROPIC_DEFAULT_SONNET_MODEL": {
+          "value": "gemini/gemini-3.0-pro"
+        },
+        "ANTHROPIC_DEFAULT_OPUS_MODEL": {
+          "value": "gemini/gemini-3.0-pro"
+        }
+      },
+      "validation": {
+        "required_env_vars": ["GEMINI_API_KEY"]
+      }
+    },
+    "litellm-openai": {
+      "description": "OpenAI GPT-4 via LiteLLM",
+      "enabled": false,
+      "env_vars": {
+        "ANTHROPIC_API_KEY": {
+          "value": "${OPENAI_API_KEY}",
+          "required": true
+        },
+        "ANTHROPIC_BASE_URL": {
+          "value": "http://localhost:4000"
+        },
+        "ANTHROPIC_DEFAULT_HAIKU_MODEL": {
+          "value": "openai/gpt-4o-mini"
+        },
+        "ANTHROPIC_DEFAULT_SONNET_MODEL": {
+          "value": "openai/gpt-4o"
+        },
+        "ANTHROPIC_DEFAULT_OPUS_MODEL": {
+          "value": "openai/gpt-4o"
+        }
+      },
+      "validation": {
+        "required_env_vars": ["OPENAI_API_KEY"]
+      }
+    },
+    "gemini-ccr": {
+      "description": "Google Gemini 2 via Claude Code Router",
+      "enabled": true,
+      "ccr_provider": true,
+      "ccr_config": {
+        "name": "gemini",
+        "api_base_url": "https://generativelanguage.googleapis.com/v1beta/models/",
+        "api_key_env": "GEMINI_API_KEY",
+        "models": {
+          "default": "gemini-2.0-flash",
+          "background": "gemini-2.0-flash",
+          "think": "gemini-2.0-flash-thinking-exp"
+        },
+        "transformer": {
+          "use": ["gemini", "enhancetool", "cleancache"]
+        }
+      },
+      "routing_modes": {
+        "standard": {
+          "default": "gemini-2.0-flash",
+          "background": "gemini-2.0-flash"
+        },
+        "reasoning": {
+          "default": "gemini-2.0-flash",
+          "think": "gemini-2.0-flash-thinking-exp",
+          "background": "gemini-2.0-flash"
+        },
+        "eco": {
+          "default": "gemini-2.0-flash",
+          "background": "gemini-2.0-flash"
+        }
+      },
+      "validation": {
+        "required_env_vars": ["GEMINI_API_KEY"]
+      }
     }
   }
 }