Configuration

The CLI stores its configuration in ~/.agent-to-bricks/config.yaml. The bricks config commands read and write this file, but you can also edit it directly.

Config file reference

Here’s a complete config file with every available option:

site:
  url: https://your-site.com
  api_key: atb_a3b7c9d2e8f4...

llm:
  provider: openai
  api_key: sk-proj-abc123...
  model: gpt-4o
  base_url: ""
  temperature: 0.3

Site settings

Key	Description	Example
site.url	Your WordPress site URL, including the protocol	https://your-site.com
site.api_key	API key from Settings > Agent to Bricks in WordPress	atb_a3b7c9d2e8f4...

LLM settings

Key	Description	Default
llm.provider	The AI provider to use	(none)
llm.api_key	API key for the provider	(none)
llm.model	Model name	(none)
llm.base_url	Custom API endpoint URL	"" (uses provider default)
llm.temperature	Controls randomness in generation (0.0 = deterministic, 1.0 = creative)	0.3

Setting values

Use bricks config set to update individual keys:

bricks config set site.url https://your-site.com
bricks config set site.api_key atb_a3b7c9d2e8f4
bricks config set llm.provider anthropic
bricks config set llm.model claude-sonnet-4-20250514

The llm.temperature value can be set by editing the config file directly at ~/.agent-to-bricks/config.yaml.

Or run the interactive wizard to set everything at once:

bricks config init

Environment variables

You can override any config value with an environment variable. This is useful for CI pipelines, Docker containers, or keeping API keys out of the config file.

The pattern is ATB_ followed by the config key in uppercase with underscores replacing dots:

Config key	Environment variable
site.url	ATB_SITE_URL
site.api_key	ATB_SITE_API_KEY
llm.provider	ATB_LLM_PROVIDER
llm.api_key	ATB_LLM_API_KEY
llm.model	ATB_LLM_MODEL
llm.base_url	ATB_LLM_BASE_URL
llm.temperature	ATB_LLM_TEMPERATURE

Environment variables take precedence over the config file. Example:

ATB_SITE_URL=https://staging.your-site.com \
ATB_SITE_API_KEY=atb_staging_key_here_here \
bricks site info

This connects to your staging site without modifying config.yaml.

LLM provider setup

The CLI works with OpenAI, Anthropic, Cerebras, and any provider that exposes an OpenAI-compatible API.

OpenAI

bricks config set llm.provider openai
bricks config set llm.api_key sk-proj-abc123...
bricks config set llm.model gpt-4o

Other OpenAI models that work well: gpt-4o-mini (faster, cheaper), gpt-4-turbo.

Anthropic

bricks config set llm.provider anthropic
bricks config set llm.api_key sk-ant-api03-abc123...
bricks config set llm.model claude-sonnet-4-20250514

Anthropic models use a different API format than OpenAI. The CLI handles the translation automatically when you set the provider to anthropic.

Cerebras

bricks config set llm.provider cerebras
bricks config set llm.api_key csk-abc123...
bricks config set llm.model llama-4-scout-17b-16e-instruct

Cerebras runs open-source models on custom hardware. Response times tend to be fast, which makes it a good option for iterating quickly.

Custom / self-hosted providers

Any provider with an OpenAI-compatible API works. Set the provider to openai and point base_url at your endpoint:

bricks config set llm.provider openai
bricks config set llm.base_url http://localhost:11434/v1
bricks config set llm.api_key ollama
bricks config set llm.model llama3.1

This works with Ollama, LM Studio, vLLM, and similar tools. The api_key field still needs a value even if your local server doesn’t require authentication — just set it to any non-empty string.

Together AI

bricks config set llm.provider openai
bricks config set llm.base_url https://api.together.xyz/v1
bricks config set llm.api_key your-together-key
bricks config set llm.model meta-llama/Llama-3-70b-chat-hf

OpenRouter

bricks config set llm.provider openai
bricks config set llm.base_url https://openrouter.ai/api/v1
bricks config set llm.api_key your-openrouter-key
bricks config set llm.model anthropic/claude-sonnet-4-20250514

Temperature

The temperature setting controls how creative or predictable the AI output is:

• 0.0 — 0.2: Very consistent output. Good for structured tasks like converting a specific layout description into HTML.
• 0.3 — 0.5: Balanced. The default of 0.3 works well for most generation tasks.
• 0.6 — 1.0: More varied output. Can produce more interesting copy but may also introduce unexpected structural choices.

To change the temperature, edit ~/.agent-to-bricks/config.yaml directly:

llm:
  temperature: 0.2

Multiple site configurations

The config file supports a single site by default. To work with multiple sites, use environment variables to switch between them:

# Production
ATB_SITE_URL=https://your-site.com \
ATB_SITE_API_KEY=atb_prod_key_here \
bricks site info

# Staging
ATB_SITE_URL=https://staging.your-site.com \
ATB_SITE_API_KEY=atb_staging_key_here \
bricks site info

For convenience, wrap these in shell aliases or a small script:

# In your .bashrc or .zshrc
alias bricks-prod='ATB_SITE_URL=https://your-site.com ATB_SITE_API_KEY=atb_prod_key_here bricks'
alias bricks-staging='ATB_SITE_URL=https://staging.your-site.com ATB_SITE_API_KEY=atb_staging_key_here bricks'

Then use them like any other command:

bricks-staging site info
bricks-prod site pull 1460

Config file location

The config file lives at ~/.agent-to-bricks/config.yaml on all platforms:

Platform	Path
Mac / Linux	~/.agent-to-bricks/config.yaml
Windows	C:\Users\YourName\.agent-to-bricks\config.yaml

The CLI creates this directory and file automatically the first time you run bricks config init or bricks config set.

Viewing current config

To see what’s currently configured:

bricks config list

This prints the active configuration with API keys partially redacted.