Providers & Models

Providers & Models

clido supports multiple LLM providers. Each provider is configured as a profile in config.toml. You can switch providers at runtime with --provider and --model.

Supported providers

Provider	Key in config	API key variable	Notes
Anthropic	anthropic	ANTHROPIC_API_KEY	Default; supports prompt caching
OpenAI	openai	OPENAI_API_KEY	Any OpenAI-API endpoint
OpenRouter	openrouter	OPENROUTER_API_KEY	Multi-model aggregator
Google Gemini	gemini	GEMINI_API_KEY
DeepSeek	deepseek	DEEPSEEK_API_KEY
Mistral	mistral	MISTRAL_API_KEY
xAI (Grok)	xai	XAI_API_KEY
Groq	groq	GROQ_API_KEY
Together AI	togetherai	TOGETHER_API_KEY
Fireworks AI	fireworks	FIREWORKS_API_KEY
Cerebras	cerebras	CEREBRAS_API_KEY
Perplexity	perplexity	PERPLEXITY_API_KEY
MiniMax	minimax	MINIMAX_API_KEY	MiniMax-M2.7 coding model; 204k context
Alibaba Cloud	alibabacloud	DASHSCOPE_API_KEY	DashScope / Qwen models
Kimi (Moonshot)	kimi	MOONSHOT_API_KEY
Kimi Code	kimi-code	KIMI_CODE_API_KEY
Local (Ollama)	local	—	No API key required

Anthropic (Claude)

The default provider. Connects to https://api.anthropic.com.

[profile.default]
provider    = "anthropic"
model       = "claude-sonnet-4-5"
api_key_env = "ANTHROPIC_API_KEY"

Recommended models

Model	Description
claude-sonnet-4-5	Best balance of capability and cost (default)
claude-3-opus-20240229	Highest capability, highest cost
claude-haiku-4-5	Fastest and cheapest; good for simple tasks

Prompt caching

When using Anthropic, clido automatically enables prompt caching for the system prompt and long conversation histories. Cache hits are billed at ~10% of the normal input token price, which significantly reduces costs for long sessions.

OpenAI-compatible

Any server that implements the OpenAI Chat Completions API can be used. This includes Azure OpenAI, Together AI, Groq, Fireworks AI, and others.

[profile.azure]
provider    = "openai"
model       = "gpt-4o"
api_key_env = "AZURE_OPENAI_API_KEY"
base_url    = "https://my-resource.openai.azure.com/openai/deployments/gpt-4o"

OpenAI

[profile.gpt4]
provider    = "openai"
model       = "gpt-4o"
api_key_env = "OPENAI_API_KEY"

OpenRouter

OpenRouter provides access to many models (Claude, GPT, Mistral, Gemini, and more) through a single API key. Models are identified by provider/model-name.

[profile.openrouter]
provider    = "openrouter"
model       = "anthropic/claude-3-5-sonnet"
api_key_env = "OPENROUTER_API_KEY"

Common OpenRouter models

Model string	Description
anthropic/claude-3-5-sonnet	Claude 3.5 Sonnet via OpenRouter
openai/gpt-4o	GPT-4o via OpenRouter
mistralai/mistral-large	Mistral Large
google/gemini-pro-1.5	Gemini Pro 1.5
meta-llama/llama-3.1-70b-instruct	Meta Llama 3.1 70B

MiniMax

Connects to https://api.minimax.io/v1. Get an API key at platform.minimax.io.

[profile.minimax]
provider    = "minimax"
model       = "MiniMax-M2.7"
api_key_env = "MINIMAX_API_KEY"

Recommended models

Model	Description
MiniMax-M2.7	Latest coding model; 204k context window
MiniMax-M1	Previous generation reasoning model

Local (Ollama)

Run models locally with Ollama. No API key or network connection required after pulling the model.

[profile.local]
provider = "local"
model    = "llama3.2"
base_url = "http://localhost:11434"

Start Ollama and pull a model:

ollama serve              # start the server
ollama pull llama3.2      # download the model

::: warning Local model limitations Local models generally have smaller context windows and weaker instruction-following than cloud models. Complex multi-step coding tasks may require a larger model (e.g. llama3.1:70b) for reliable tool use. :::

Alibaba Cloud (DashScope / Qwen)

Connects to the DashScope OpenAI-compatible endpoint. Set DASHSCOPE_API_KEY or store the key in your profile.

[profile.alibaba]
provider    = "alibabacloud"
model       = "qwen-max"
api_key_env = "DASHSCOPE_API_KEY"

You can override the endpoint with base_url if needed (defaults to https://dashscope.aliyuncs.com/compatible-mode/v1).

Listing available models

List all models known to clido for a provider (from the built-in pricing table):

clido list-models
clido list-models --provider anthropic
clido list-models --provider openrouter --json

Fetching the current model list from a provider’s API

Retrieve the live model list from a provider:

clido fetch-models
clido fetch-models --provider openrouter

::: tip Updating pricing data If model pricing changes, update the local pricing table:

clido update-pricing

:::

Switching provider and model at runtime

Override the profile’s provider and model for a single run:

# Use a different model
clido --model claude-haiku-4-5 "quick task"

# Use a different profile
clido --profile local "quick task"

# Override both
clido --provider openrouter --model anthropic/claude-3-5-sonnet "task"

These flags only affect the current invocation; they do not modify config.toml.

Per-profile base_url

For providers that need a custom endpoint (Azure, self-hosted, Ollama):

[profile.custom]
provider = "openai"
model    = "my-model"
base_url = "https://my-server.internal/v1"
api_key_env = "MY_API_KEY"

The base_url is used as-is; clido appends /chat/completions (OpenAI-compatible providers) or the appropriate path for each provider.

API key security

clido looks for API keys in this order:

1. Environment variable (e.g. ANTHROPIC_API_KEY)
2. Credentials file (~/.config/clido/credentials, created automatically during setup with chmod 600 permissions)
3. Environment variable named by api_key_env in the profile
4. api_key field in the profile (legacy fallback — not recommended)

The recommended approach is to let the setup wizard store keys in the credentials file, or set the environment variable in your shell profile:

# ~/.zshrc or ~/.bashrc
export ANTHROPIC_API_KEY="sk-ant-..."