API Keys & Auth
gptcgt offers two ways to power its AI models. You choose which approach fits your workflow — or use both.
Option 1: Bring Your Own Keys (BYOK) — Free
Supply your own API keys from any supported provider. gptcgt stores them securely in your operating system's native keychain (macOS Keychain, Windows Credential Locker, Linux Secret Service) via the keyring Python library. Keys never touch disk in plaintext.
Supported Providers
| Provider | Env Variable | Notable Models |
|---|---|---|
| OpenAI | OPENAI_API_KEY | GPT-4o, o3, o1 |
| Anthropic | ANTHROPIC_API_KEY | Claude 3.5 Sonnet, Claude 3 Opus |
| GEMINI_API_KEY | Gemini 2.5 Pro, Gemini 2.5 Flash | |
| DeepSeek | DEEPSEEK_API_KEY | DeepSeek V3, DeepSeek Coder |
| xAI | XAI_API_KEY | Grok 2 |
| Mistral | MISTRAL_API_KEY | Mistral Large |
| Groq | GROQ_API_KEY | LLaMA 3 (ultra-fast inference) |
| Cohere | COHERE_API_KEY | Command R+ |
| OpenRouter | OPENROUTER_API_KEY | Access to 200+ models via one key |
| Custom | CUSTOM_API_KEY | Ollama, vLLM, any OpenAI-compatible API |
Pros: Only pay providers directly. Total privacy. No subscription needed.
Cons: You manage rate limits and billing across multiple providers yourself. Ensemble/Architect modes require keys from multiple providers.
BYOK Safety Limits
Even with your own keys, gptcgt enforces spend protection:
# In ~/.gptcgt/global.toml
daily_spend_limit = 10.0 # $10/day hard stop
max_spend_per_task = 2.0 # $2 max per individual task
max_tokens_per_task = 500000 # 500K token cap per taskOption 2: Managed Credits (Subscription)
Sign up for gptcgt Pro and get a single bucket of Credits that work seamlessly across all providers. We handle API keys, routing, rate limiting, and billing through our LiteLLM proxy.
- Type
/loginin the app to start the WorkOS device flow - Authorize in your browser
- You're in — credits are automatically used for all tasks
Credit Costs by Mode
| Mode | Credits | ~ Dollar Cost |
|---|---|---|
| Scout | 1 | ~$0.01 |
| Standard | 5 | ~$0.05 |
| Ensemble / Battle | 25 | ~$0.25 |
| Architect | 100 | ~$1.00 |
Key Commands
/login— Authenticate with your gptcgt account/logout— Sign out/credits— Check your remaining balance/billing— Open the web dashboard for subscription management
Verifying Key Health
Run /status at any time to check the health and latency of all configured providers. gptcgt will test each API key with a minimal health-check request using the cheapest available model for that provider.