Providers

Astreus CLI supports multiple LLM providers. Switch between them seamlessly to test your agents with different models and find the best fit for your use case.

Supported Providers#

Examples select models explicitly; existing supported defaults are unchanged. Retired native models need an explicit current replacement.

The existing CLI reads model choices from the installed framework via getModelsByProvider. The four choices remain openai, claude, gemini, and ollama.

OpenAI#

The default provider with access to GPT models.

bash

export OPENAI_API_KEY="${OPENAI_API_KEY:?Set OPENAI_API_KEY in your environment}"
astreus

Model IDs to select explicitly:

gpt-6-astra
gpt-5.6-sol
gpt-5.6-terra
gpt-5.6-luna

Hosted free models through OpenRouter#

Keep ASTREUS_PROVIDER=openai for hosted free models. Supply an OpenRouter key through the environment, and select the model explicitly. The same rate limits and endpoint requirements apply as in the framework LLM guide.

bash

export OPENAI_BASE_URL="https://openrouter.ai/api/v1"
export OPENAI_API_KEY="${OPENROUTER_API_KEY:?Set OPENROUTER_API_KEY in your environment}"
ASTREUS_PROVIDER=openai ASTREUS_MODEL=openrouter/free astreus

Anthropic (Claude)#

Access Claude models for nuanced conversations.

bash

export ANTHROPIC_API_KEY="${ANTHROPIC_API_KEY:?Set ANTHROPIC_API_KEY in your environment}"
ASTREUS_PROVIDER=claude ASTREUS_MODEL=claude-sonnet-5 astreus

Model IDs to select explicitly:

claude-fable-5-1
claude-opus-5
claude-sonnet-5
claude-haiku-4-5-20251001

Google (Gemini)#

Use Google's Gemini models.

bash

export GEMINI_API_KEY="${GEMINI_API_KEY:?Set GEMINI_API_KEY in your environment}"
ASTREUS_PROVIDER=gemini astreus

Available Models:

gemini-pro (default) - Gemini Pro
gemini-pro-vision - Vision capable

Ollama (Local Models)#

Run models locally with Ollama. No API key required.

bash

# Make sure Ollama is running
ollama serve

# Start CLI with Ollama
ASTREUS_PROVIDER=ollama astreus

Available Models:

llama3 (default) - Llama 3
llama2 - Llama 2
mistral - Mistral
codellama - Code-focused Llama

Switching Providers#

At Startup#

Use environment variables:

bash

ASTREUS_PROVIDER=claude ASTREUS_MODEL=claude-sonnet-5 astreus

During Session#

Use slash commands to switch:

text

/provider claude
/model claude-sonnet-5

Or open the interactive selector:

text

/provider
/model

Provider Configuration#

Custom Endpoints#

For enterprise deployments or proxies:

bash

# Ollama on different host
export OLLAMA_HOST=http://192.168.1.100:11434

Vision Support#

Some providers support vision (image) capabilities:

bash

# Separate API key for vision (optional)
export OPENAI_VISION_API_KEY=sk-your-vision-key
export ANTHROPIC_VISION_API_KEY=your-vision-key
export GEMINI_VISION_API_KEY=your-vision-key

Embedding Support#

For memory and knowledge features:

bash

export OPENAI_EMBEDDING_API_KEY=sk-your-embedding-key
export GEMINI_EMBEDDING_API_KEY=your-embedding-key

Model Selection for Agent Development#

Choose the right model based on the type of agent you're building:

Agent Type	Recommended Model	Why
Code Analysis Agent	`claude-sonnet-5`	Excellent at understanding code structure and patterns
Complex Reasoning Agent	`gpt-4o`, `claude-sonnet-5`	Best for multi-step reasoning and planning
Fast Response Agent	`gpt-4o-mini`, `gemini-pro`	Low latency for real-time interactions
Tool-Heavy Agent	`gpt-4o`	Reliable function calling and tool use
Long Context Agent	`claude-sonnet-5`	Check the selected model and endpoint’s context limit before sending large documents
Local/Private Agent	`llama3`, `codellama`	No data leaves your machine
Cost-Sensitive Agent	`gpt-3.5-turbo`, Ollama	Minimize API costs

Testing Agents Across Providers#

When building agents, it's useful to test with different providers:

text

# Start developing with fast, cheap model
/provider openai
/model gpt-4o-mini

# Test complex reasoning with top models
/provider claude
/model claude-sonnet-5

# Test locally before deployment
/provider ollama
/model llama3

This workflow helps you understand how your agent performs across different LLMs and find the best balance of capability, speed, and cost for your use case.