Supported Providers
| Provider | Models | API Key Required |
|---|---|---|
| Anthropic | Claude Opus, Sonnet, Haiku | Yes (bring your own) |
| OpenAI | GPT-5.2, GPT-5, o3, o4 | Yes (bring your own) |
| OpenRouter | 100+ models | Yes (bring your own) |
Using Different Models
Anthropic (Default)
OpenAI
Use your own OpenAI API key for GPT models:Add your OpenAI API key in the project settings at app.chucky.cloud.
OpenRouter
Access 100+ models through OpenRouter using theor: prefix:
Add your OpenRouter API key in the project settings. Get one at openrouter.ai.
Model Detection
Chucky automatically routes requests based on the model name:| Pattern | Provider |
|---|---|
or:* | OpenRouter (explicit) |
claude-* | Anthropic |
gpt-*, o1*, o3*, o4* | OpenAI |
| Everything else | OpenRouter (fallback) |
Examples
Environment Variables
Configure Claude Code model behavior via environment variables passed to sessions:Primary Model
| Variable | Description |
|---|---|
ANTHROPIC_MODEL | Primary model for the session |
Model Aliases
Override the default models for each alias tier:| Variable | Description | Default |
|---|---|---|
ANTHROPIC_DEFAULT_OPUS_MODEL | Model for opus alias | claude-opus-4-5-20251101 |
ANTHROPIC_DEFAULT_SONNET_MODEL | Model for sonnet alias | claude-sonnet-4-5-20250929 |
ANTHROPIC_DEFAULT_HAIKU_MODEL | Model for haiku alias and background tasks | claude-haiku-4-5-20251001 |
Subagent Model
| Variable | Description |
|---|---|
CLAUDE_CODE_SUBAGENT_MODEL | Model used for subagent/background tasks |
Prompt Caching
Control prompt caching behavior:| Variable | Description |
|---|---|
DISABLE_PROMPT_CACHING | Set to 1 to disable all caching |
DISABLE_PROMPT_CACHING_HAIKU | Set to 1 to disable for Haiku |
DISABLE_PROMPT_CACHING_SONNET | Set to 1 to disable for Sonnet |
DISABLE_PROMPT_CACHING_OPUS | Set to 1 to disable for Opus |
These environment variables are passed to the Claude Code sandbox and affect how models are selected for different tasks within a session.
Custom Pricing
For OpenRouter and other non-standard models, you can configure custom pricing to ensure accurate cost tracking.Why Custom Pricing?
- OpenRouter models have varying prices
- Some models are free (e.g.,
mistralai/devstral-2512:free) - Accurate billing for your users
Setting Up Custom Pricing
- Go to Billing → Custom Pricing in the portal
- Add pricing entries for models you use
| Field | Description | Example |
|---|---|---|
| Model Pattern | Exact name or wildcard | mistralai/* |
| Input Price | USD per 1M input tokens | 2.00 |
| Output Price | USD per 1M output tokens | 6.00 |
Wildcard Patterns
Use/* suffix to match all models from a provider:
Pricing Priority
When calculating costs, Chucky checks in order:- Custom exact match - Your pricing for the exact model name
- Custom wildcard - Your pricing matching
provider/* - Built-in pricing - Known Anthropic/OpenAI prices
- Default fallback - Sonnet-level pricing (15 per 1M tokens)
Example: Free Models
For free OpenRouter models:| Model Pattern | Input Price | Output Price |
|---|---|---|
mistralai/devstral-2512:free | 0 | 0 |
meta-llama/llama-3-8b:free | 0 | 0 |
Provider Configuration
Adding API Keys
In the portal at Projects → [Your Project] → Settings:1
Anthropic API Key
Required for Claude models. Get one at console.anthropic.com.
2
OpenAI API Key
Required for GPT models. Get one at platform.openai.com.
3
OpenRouter API Key
Required for OpenRouter models. Get one at openrouter.ai.
Custom OpenRouter Endpoint
For self-hosted OpenRouter or compatible APIs:Model Recommendations
| Use Case | Recommended Model | Why |
|---|---|---|
| General coding | claude-sonnet-4-5-20250929 | Best balance of speed/quality |
| Complex reasoning | claude-opus-4-5-20251101 | Highest capability |
| High volume | claude-3-5-haiku-20241022 | Fastest, cheapest |
| Cost-sensitive | or:deepseek/deepseek-chat | Very cheap, good quality |
| Free tier | or:mistralai/devstral-2512:free | No cost |