Skip to content

Providers

Vinh Nguyen edited this page Aug 5, 2026 · 1 revision

Providers

VT Code supports 21+ LLM providers out of the box, plus any OpenAI-compatible API via [[custom_providers]].

Provider Categories

Category Providers
Cloud LLMs Anthropic (Claude), OpenAI (GPT), Google Gemini, Z.AI, Moonshot (Kimi), StepFun, MiniMax, Mistral, Qwen
Gateways OpenRouter, Evolink, HuggingFace, Atlas Cloud, OmniRoute
Local inference Ollama, LM Studio, llama.cpp
Other GitHub Copilot, Anthropic API Compat, Poolside

Setting an API Key

Use the OS keyring (recommended):

/secret add openai        # from the TUI
vtcode secret add gemini  # from the CLI

Or export environment variables:

export ANTHROPIC_API_KEY=...
export GEMINI_API_KEY=...
export OPENAI_API_KEY=...

Selecting a Model

Run /model inside a session to open the model picker. The picker respects providers_whitelist when set.

agent.reasoning_effort controls reasoning depth for models that support it (none, minimal, low, medium, high, xhigh, max) — applies to Claude, GPT-5 family, Gemini, Qwen3, and DeepSeek with reasoning capability.

Custom Providers

Add any OpenAI-compatible endpoint (private gateway, aggregator, or internal inference cluster):

[[custom_providers]]
name = "mycorp"
display_name = "MyCorp"
base_url = "https://llm.corp.example/v1"
api_key_env = "MYCORP_API_KEY"
model = "gpt-5-mini"
context_window = 256000   # optional; defaults to 128000 tokens

Custom provider entries are editable in /config and appear in the model picker using display_name. Declare models = [...] to offer several models from one endpoint (handy for aggregators like Atlas Cloud).

Provider Governance

Use providers_whitelist to restrict which providers VT Code may access — important in corporate or air-gapped environments to prevent accidental data leakage to unapproved endpoints:

providers_whitelist = ["opencode-zen", "opencode-go", "gemini"]

When set, the /model picker, first-run wizard, and config validator all respect the list. Empty (default) allows everything.

Auth Options

  • OAuth flows for OpenRouter and OpenAI (auth.openrouter.use_oauth, auth.openai.preferred_method).
  • GitHub Copilot auth via auth.copilot.* (host, command, tool allow-list).
  • agent.credential_storage_mode: keyring (default), file (AES-256-GCM with machine-derived key), or auto.

Related

Clone this wiki locally