opencode-nexos-models-config

Simple CLI tool that generates opencode configuration for Nexos AI models. It fetches available models from the Nexos AI API and creates an opencode configuration file with proper model limits, costs, and variants.

Requirements

• Node.js 18+
• opencode installed globally (npm install -g opencode)
• NEXOS_API_KEY environment variable set with your Nexos AI API key

Installation

Global installation (recommended)

npm install -g opencode-nexos-models-config

Per-user installation (without root/sudo)

Linux

Add to ~/.npmrc:

prefix=~/.npm-global

Add to ~/.bashrc:

export PATH="$HOME/.npm-global/bin:$PATH"

Then reload shell and install:

source ~/.bashrc
npm install -g opencode-nexos-models-config

macOS

Add to ~/.npmrc:

prefix=~/.npm-global

Add to ~/.zshrc:

export PATH="$HOME/.npm-global/bin:$PATH"

Then reload shell and install:

source ~/.zshrc
npm install -g opencode-nexos-models-config

Updating

npm update -g opencode-nexos-models-config

Or to force the latest version:

npm install -g opencode-nexos-models-config@latest

Usage

After global installation:

opencode-nexos-models-config

Or without installation (using npx):

npx opencode-nexos-models-config

The tool will:

1. Fetch the list of available models from the Nexos AI API
2. Generate an opencode configuration with the nexos-provider plugin
3. Write the config to ~/.config/opencode/opencode.json

Supported models only

By default, only models with predefined configuration (costs, limits, variants) are included. This is equivalent to running:

opencode-nexos-models-config
# or
opencode-nexos-models-config --supported-models
# or
opencode-nexos-models-config -m

To include all available models from the API (not just supported ones):

opencode-nexos-models-config --supported-models=false
# or
opencode-nexos-models-config -m false

This filters the list to only include models from the curated SUPPORTED_MODELS list (see below).

Agent model selection

If you have agents defined in your opencode.json (e.g. build, build-fast, build-heavy, plan), you can interactively assign models to them:

opencode-nexos-models-config --select-agents

This opens an interactive prompt for each agent where you can search and select a model using arrow keys and type-to-filter.

Custom model costs

To interactively set custom prices for models in your config:

opencode-nexos-models-config --custom-costs

Or use the short flag:

opencode-nexos-models-config -c

This allows you to:

• Set custom input/output/cache prices per model
• See current costs while selecting
• Update prices for multiple models in one session
• Leave fields blank to keep existing values

When used with --supported-models, only supported models will be shown:

opencode-nexos-models-config -m -c

Model pricing information

The tool automatically includes pricing information for all models in the generated configuration. Pricing includes:

• Input cost: Price per million input tokens
• Output cost: Price per million output tokens
• Cache read: Price per million cached tokens read (if supported)
• Cache write: Price per million tokens written to cache (if supported)

Default pricing:

• Supported models have predefined costs (see table below)
• Unknown models use fallback costs based on Claude Opus 4.6 pricing:
  • Input: $5/M tokens
  • Output: $25/M tokens
  • Cache read: $0.5/M tokens
  • Cache write: $6.25/M tokens

Custom pricing:

• Your existing custom costs are always preserved
• Use --custom-costs flag to set new custom prices interactively
• Custom costs take priority over default pricing

Note: The app always overwrites the entire model list with fresh data from the API on each run.

Configuration

Environment Variables

Variable	Description	Default
NEXOS_API_KEY	Your Nexos AI API key (required)	-
NEXOS_BASE_URL	Custom API base URL	https://api.nexos.ai/v1

CLI Flags

Flag	Description	Default
--help, -h	Show help message	-
--version, -v	Show version number	-
--select-agents, -s	Interactively select models for agents defined in config	false
--supported-models, -m	Only include models with predefined configuration (true/false)	true
--custom-costs, -c	Interactively set custom costs for models	false
--output, -o	Write config to a custom file path instead of default

Supported Models

When using --supported-models flag, only these curated models are included:

Model	Provider	Context	Output	Variants
Claude Opus 4.5	Anthropic	200k	64k	low, high
Claude Opus 4.6	Anthropic	200k	128k	low, high
Claude Sonnet 4.5	Anthropic	200k	64k	low, high
GPT 5.2	OpenAI	400k	128k	low, high
GPT 5	OpenAI	400k	128k	low, high
Gemini 2.5 Pro	Google	1M	65k	low, high
Gemini 2.5 Flash	Google	1M	65k	low, high
Kimi K2.5	Moonshot AI	256k	64k	-

All models come with pre-configured context limits, output limits, and pricing information.

Variants

• low: Minimal thinking/reasoning (faster, cheaper)
• high: Extended thinking/reasoning (more accurate for complex tasks)

Default mode

By default, the tool only includes supported models from the curated list above. This ensures you work with models that have pre-configured costs, limits, and variants.

To include all available models from the Nexos AI API (including unsupported ones):

opencode-nexos-models-config --supported-models=false

The entire model list is replaced on each run to ensure you have the latest models and pricing.

Important: Your custom costs set via --custom-costs are preserved and take priority, but the model list itself is always refreshed from the API.

Model Configuration

Model metadata is stored in models.config.mjs and includes:

• limits: Context window and max output tokens
• cost: Input/output pricing per 1M tokens, plus cache read/write if supported
• variants: Model-specific thinking/reasoning variants (low, high)
• options: Default model options (e.g., reasoningEffort)

License

MIT