Getting Started with CLI - Interactive Configuration

Updated on Oct 1, 2026

The interactive setup is the recommended way to configure Forge for a project. It guides you through choosing an AI coding agent, reviewing AI-generated project metadata, connecting development tools, and confirming the files before Forge writes them.

Prerequisites

Before you begin:

Install Forge and make sure frg --version succeeds. See Installation.
Open an interactive terminal at the root of your project.
Install and authenticate at least one supported AI coding agent. Forge uses that agent to analyze the repository.
Use a Git repository when you want Forge to detect repository and branch information or run source-control workflows.

You can run interactive configuration without a Telerik license. A license is required when you start agent-backed workflow commands. See Licensing.

Start the Setup

Run the following command from the project root:

bash

frg config init -i

The setup has four phases. At each prompt, select an option or enter a custom value. Press Enter to accept the displayed default or suggestion. Enter back to return to the previous prompt or quit to cancel the setup.

Phase 1: Configure the AI Agent

Select the AI coding agent that Forge will use. The available choices include GitHub Copilot CLI, Claude Code, and OpenCode.

Next, select a model profile:

Profile	Use case
`lite`	Lower-cost experimentation
`balanced`	Balance cost and capability
`stable`	Predictable behavior with stable models
`heavy`	Maximum capability for complex work

The wizard displays the models selected by the profile. Press Enter to keep each default or enter another model when you need an override.

If you skip agent configuration, Forge cannot analyze the repository and asks you to enter the project metadata manually.

Phase 2: Review Project Metadata

Forge uses the selected agent to inspect the repository and suggest these required values:

Project name and description
Primary programming language and framework
Testing framework
Package manager

Review every suggestion. Press Enter to accept it or type a replacement. For a detected monorepo, also review the suggested services and their paths.

If repository analysis fails, retry it, enter the metadata manually, or quit without saving the configuration.

Phase 3: Connect Development Tools

Select the tools that Forge workflows can use:

Issue tracker
Code platform
Ticket tracker, if required
Security scanner, if required

Optional integrations can be skipped. When you select an issue or ticket tracker, the wizard also asks for the required code platform. Forge can detect repository and base-branch information from Git, and you can accept or replace the suggestions.

By default, Forge configures supported integrations through generated agent skills. To choose another access method, such as a CLI, MCP server, or API, start the wizard with:

bash

frg config init -i --advanced

Phase 4: Confirm the Configuration

Review the summary of your agent, project metadata, tool integrations, and files to be written. You can:

Write the configuration.
Return to one phase and change its answers while keeping the others.
Cancel without writing the configuration.

Depending on your selections, Forge creates:

.forge/config/project.toml for project metadata.
.forge/config/agents.toml for the selected coding agent and models.
.forge/config/toolchain.toml for development tool integrations.
Generated skill files for integrations that use the skill access method.

When configuration files already exist, the wizard shows what will be preserved or replaced and asks for confirmation before overwriting them.

Validate the Setup

After the wizard finishes, validate the generated configuration and check the environment:

bash

frg config validate
frg status

Resolve any reported authentication, executable, or configuration errors before running a workflow.

Run Setup Non-Interactively

Interactive setup requires a terminal and cannot read answers from piped input. For scripts and CI, pass the required options explicitly instead:

bash

frg config init --agent github_copilot --models stable \
  --issues github_issues --code github

See Project Configuration for all initialization options and advanced configuration fields.

Run Your First Workflow

Set up your Telerik license if you have not already. Set the context to an existing issue in the tracker selected during setup. Replace <ISSUE_ID> with its identifier:

bash

frg config set-issue <ISSUE_ID>
frg workflow run --auto

With --auto, Forge classifies the configured issue and selects a suitable workflow. Review the selection and confirm it when prompted. Forge then runs that workflow. Its steps depend on the workflow selected, and it may pause at approval gates for your review.

Next Steps

Follow the workflow's progress and respond to approval or input prompts when they appear.
Review the workflow results and generated artifacts. If it creates a pull request, review its changes and checks in your code platform before merging.
To run Auto for another issue, set that issue as the context with frg config set-issue <ISSUE_ID>, then run frg workflow run --auto again.
To choose a workflow yourself instead of using Auto, run frg workflow run <WORKFLOW_NAME>.