​
Using Commands in Chat

Quickly access any configured command in chat: Commands dropdown showing available commands when typing forward slash Commands are designed to:
  • automate common repetitive tasks, with minimal prompting
  • help your team to enforce standards around a task (ie. generating an RFC, changelog update, etc.)
  • be extensible (set the model, mode)

​
Cookbooks

Ready-to-use command setups with complete firebender.json configurations:

​
Configuration

Commands are configured in your firebender.json file using the commands array: 
{
  "commands": [
    {
      "name": "create pr",
      "path": "~/firebender-commands/pr-description.md",
      "model": "gpt-4 128k",
      "mode": "write"
    },
    {
      "name": "add integration test for git diff",
      "path": "./prompts/add-integration-tests.md",
      "mode": "auto"
      // not providing model, will default to the one you have picked
    },
    {
      "name": "review code",
      "path": "./prompts/code-review.md",
      "model": "gpt-3.5 turbo",
      "mode": "read"
    }
  ]
}
The content of this file ~/firebender-commands/pr-description.md: 
### Create a GitHub Pull Request from any repo state

Create a PR using `gh` and `git` commands. Determine the best PR title and body based on other merged commits.

Make sure non default branch is used (ie. do not push to main).

If changes are uncommitted, then commit them and push.

Use `--no-pager` with git commands because you won't have access to STDIN while the command is still running. Avoiding interactive editors is best practice here.

If the current state of the repo makes it impractical to create a PR, then just plainly state why that is.
Each command requires:
  • name: The display name that appears in the UI
  • path: Path to the markdown file containing the command instructions
  • model (optional): Preferred model to use for this command
  • mode (optional): Preferred mode to use for this command

​
Model Property

The model property allows you to specify which AI model should be used when executing a specific command. This is useful for optimizing performance based on the task: Predefined Values:
  • "default" - Uses the system’s default model selection
  • "quick" - Uses the fastest available model for quick tasks
You can view all available models on our models page. You can specify any available model by its name, for example:
  • "claude-4 sonnet"
  • "gpt-5"
  • "gpt-4 128k"
  • "gpt-3.5 turbo"
  • "quick" (see above)

​
Mode Property

The mode property allows you to specify which interaction mode should be used when executing a specific command:
  • "auto" - Let Firebender automatically determine the best mode based on context (default)
  • "read" - Read-only mode for analysis, reviews, and explanations
  • "write" - Agent mode for making code changes and running terminal commands
  • "composer" - UI composition mode for creating interfaces
If the model or mode properties are omitted, the command will use whatever model and mode the user has currently selected in the chat interface.

​
Command Files

Command files are markdown documents that contain prompts or instructions for the AI. They can include:
  • Natural language instructions
  • Code examples or templates
  • Specific formatting requirements
  • Context about the codebase or project

​
Path Resolution

Command file paths are resolved based on the configuration location:
  • Project configuration: Relative paths resolve from the project root
  • Personal configuration: Relative paths resolve from ~/.firebender/
You can also use:
  • Absolute paths: /full/path/to/command.md
  • Home directory expansion: ~/commands/my-command.md