Skip to main content

Provider Commands

Manage LLM providers and their configurations. Providers are API endpoints that serve language models.

Command: lc providers (alias: lc p)​

Subcommands​

Add Provider​

Add a new provider to your configuration.

lc providers add <name> <endpoint> [OPTIONS]
lc p a <name> <endpoint> [OPTIONS]

Options:

  • -m, --models-path <PATH> - Custom models endpoint (default: /models)
  • -c, --chat-path <PATH> - Custom chat endpoint (default: /chat/completions)

Examples:

# Standard OpenAI-compatible provider
lc providers add openai https://api.openai.com/v1

# Provider with custom endpoints
lc providers add github https://models.github.ai \
  --models-path /catalog/models \
  --chat-path /inference/chat/completions

# Short form
lc p a together https://api.together.xyz/v1

List Providers​

Show all configured providers.

lc providers list
lc p l

Output example:

Configured providers:
  • openai (https://api.openai.com/v1) [key set]
  • claude (https://api.anthropic.com/v1) [key set]
  • together (https://api.together.xyz/v1) [no key]

List Models​

Show available models from a specific provider.

lc providers models <provider>
lc p m <provider>

Example:

lc providers models openai
# Output:
# Available models for openai:
#   • gpt-4-turbo-preview
#   • gpt-4
#   • gpt-3.5-turbo
#   • text-embedding-3-small
#   • text-embedding-3-large

Update Provider​

Update a provider's endpoint URL.

lc providers update <name> <endpoint>
lc p u <name> <endpoint>

Example:

lc providers update openai https://api.openai.com/v1

Manage Headers​

Add, list, or delete custom headers for providers.

Add Header

lc providers headers <provider> add <header> <value>
lc p h <provider> a <header> <value>

List Headers

lc providers headers <provider> list
lc p h <provider> l

Delete Header

lc providers headers <provider> delete <header>
lc p h <provider> d <header>

Set Token URL​

Configure token URLs for providers requiring different endpoints for token handling.

lc providers token-url <provider> <url>
lc p t <provider> <url>

Remove Provider​

Remove a provider from your configuration.

lc providers remove <name>
lc p r <name>

Example:

lc providers remove old-provider

Example: Token URL Setup

# Set a custom token URL for a provider
lc providers token-url custom-provider https://api.custom.com/auth/token

Custom Headers​

Some providers require additional headers beyond the standard Authorization header.

Example: Anthropic Claude Setup

# Add Claude provider
lc providers add claude https://api.anthropic.com/v1 -c /messages

# Add required headers
lc providers headers claude add x-api-key sk-ant-api03-...
lc providers headers claude add anthropic-version 2023-06-01

# Verify headers
lc providers headers claude list

Common Provider Configurations​

OpenAI​

lc providers add openai https://api.openai.com/v1
lc keys add openai

Anthropic Claude​

lc providers add claude https://api.anthropic.com/v1 -c /messages
lc providers headers claude add x-api-key <your-key>
lc providers headers claude add anthropic-version 2023-06-01

OpenRouter​

lc providers add openrouter https://openrouter.ai/api/v1
lc keys add openrouter

Together AI​

lc providers add together https://api.together.xyz/v1
lc keys add together

GitHub Models​

lc providers add github https://models.github.ai \
  -m /catalog/models \
  -c /inference/chat/completions
lc keys add github

Local Ollama​

lc providers add ollama http://localhost:11434/v1
# No API key needed for local providers

Hugging Face Router​

lc providers add hf https://router.huggingface.co/v1
lc keys add hf

Google Vertex AI (Service Account JWT)​

Vertex AI on Google Cloud uses OAuth 2.0 with a Service Account (SA). lc supports first-class auth using the JWT Bearer flow with automatic token mint/refresh and path templating for project/location/model.

Quickstart

# 1) Add Vertex AI provider (endpoint auto-detects google_sa_jwt)
lc providers add vertex_google https://aiplatform.googleapis.com \
  -c /v1/projects/{project}/locations/{location}/publishers/google/models/{model}:generateContent

# 2) Provide project/location via provider vars
lc providers vars vertex_google set project <your-project-id>
lc providers vars vertex_google set location <your-location>   # e.g., us-central1 or global

# 3) Add Service Account JSON (paste as base64; stored encrypted)
lc keys add vertex_google
# When prompted, paste the base64 version: cat sa.json | base64

# 4) (Optional) Override token URL (defaults to https://oauth2.googleapis.com/token)
lc providers token-url vertex_google https://oauth2.googleapis.com/token

# 5) Use a Vertex model
lc -p vertex_google -m gemini-2.5-pro "Hello from Vertex"

Notes

  • Service Account JSON must minimally include:
    • type=service_account
    • client_email
    • private_key
  • lc mints an RS256-signed JWT with claims:
  • lc exchanges the assertion at the token URL for an access_token, caches it with a safety skew, and automatically refreshes when needed.
  • The chat path templates:
    • {project}, {location} from provider vars
    • {model} from the runtime -m flag
  • For Gemini API (non-Vertex) providers using x-goog-api-key, continue to use standard API key flows. Vertex AI flows use Bearer tokens obtained via the SA JWT exchange.

Troubleshooting

  • "Missing provider vars"
    • Set vars: lc providers vars vertex_google set project <id>; lc providers vars vertex_google set location <loc>
    • List vars: lc providers vars vertex_google list
  • "Invalid service account JSON" or "Invalid base64 format"
    • Re-run: lc keys add vertex_google and paste the base64 version: cat sa.json | base64
  • "Authentication failed"
    • Ensure the Service Account has Vertex AI permissions (e.g., Vertex AI User) and the project/location are correct
    • If using a VPC-SC or restricted org policy, confirm token audience and scopes are permitted

Provider Features​

Custom Endpoints​

Some providers use non-standard paths for their endpoints:

  • Models Path: Where to fetch available models (default: /models)
  • Chat Path: Where to send chat requests (default: /chat/completions)

Response Format Support​

LLM Client automatically detects and handles multiple response formats:

  1. OpenAI Format (most providers)
  2. Llama API Format (Meta)
  3. Cohere Format
  4. Anthropic Format (Claude)

Special Provider: Hugging Face Router​

The HF router expands models with their available providers:

lc providers models hf
# Output shows:
# • Qwen/Qwen3-32B:groq
# • Qwen/Qwen3-32B:hyperbolic
# • meta-llama/Llama-3.3-70B-Instruct:together

Use the full model:provider format when prompting:

lc --provider hf -m "Qwen/Qwen3-32B:groq" "Hello"

Troubleshooting​

"Provider not found"​

  • Check spelling: lc providers list
  • Ensure provider is added: lc providers add <name> <url>

"Invalid endpoint"​

  • Verify URL includes protocol: https:// or http://
  • Check if custom paths are needed: -m and -c flags

"Authentication failed"​

  • Verify API key: lc keys add <provider>
  • Check if custom headers are needed
  • Some providers use x-api-key instead of Authorization

See Also​

Next Steps​