Skip to main content

​
General Questions

Forge is an AI-enhanced terminal development environment that integrates AI capabilities with your development workflow. It’s a comprehensive coding agent that helps you write code, debug issues, refactor projects, and learn new technologies without leaving your terminal.
Yes! Forge is open-source software under the Apache 2.0 license. You can view the source code, contribute, and even fork the project on GitHub.
Forge supports multiple programming languages through its language-aware parsing and analysis system. It works with popular languages including JavaScript/TypeScript, Python, Rust, Go, Java, C/C++, and many others.
Yes, your code stays on your local machine. Forge is designed with security in mind. Only the specific code context needed for AI processing is sent to your chosen AI provider. Your entire codebase is never uploaded.
Forge is a comprehensive terminal-based development environment with multi-agent workflows, conversation management, and extensive tool integration. GitHub Copilot is primarily an IDE plugin for code completion. Forge offers deeper code understanding, project-wide operations, shell integration, and autonomous task execution.

​
Installation & Setup

The easiest way to install Forge is using the installation script:
curl -fsSL https://forgecode.dev/cli | sh
Alternatively, you can install via npm:
npm install -g forgecode
Or build from source:
git clone https://github.com/antinomyhq/forge
cd forge
cargo build --release
  • Rust: 1.75+ (for building from source)
  • Operating System: macOS, Linux, or Windows (WSL recommended)
  • Node.js: 16+ (for npm installation)
  • Git: For version control operations
  • Optional: GitHub CLI (gh) for GitHub integration
Yes, Forge requires an API key from one of the supported AI providers:
  • OpenAI
  • Anthropic
  • OpenRouter
  • Google Vertex AI
  • Amazon Bedrock
  • Cerebras
  • x-ai
  • z.ai
  • Requesty
  • IO Intelligence
  • Groq
  • OpenAI-compatible providers
Use forge provider login to configure your credentials interactively.
Forge checks for updates automatically in the background. When a new version is available:
  • You’ll see a notification
  • Run the same installation command to update:
curl -fsSL https://forgecode.dev/cli | sh
Or via npm:
npm update -g forgecode

​
Configuration

Forge uses two configuration files:
  • forge.yaml: Project-specific configuration in your project root
  • .mcp.json: MCP server configuration (local or user-level)
  • Provider credentials: Stored securely in your system’s config directory
Provider credentials are stored at:
  • macOS/Linux: ~/.config/forge/
  • Windows: %APPDATA%\forge\
You can configure multiple providers using forge provider login and switch between them:
# Add providers
forge provider login  # Follow prompts for each provider

# List providers
forge provider list

# Switch provider and model
forge --provider anthropic --model claude-3.7-sonnet
Or set defaults in forge.yaml:
model: claude-3.7-sonnet
Custom rules are guidelines you can add to forge.yaml that all AI agents follow when generating responses:
custom_rules: |
  1. Always add comprehensive error handling
  2. Include unit tests for all new functions
  3. Use TypeScript strict mode
  4. Follow our team's naming conventions
This ensures consistent code generation aligned with your team’s standards.
No, Forge requires an internet connection to communicate with AI providers. However, you can:
  • Use a local AI model with an OpenAI-compatible API server
  • Configure a self-hosted AI provider
  • Use on-premises AI services like Amazon Bedrock in your VPC

​
Usage

Simply run:
forge
This starts an interactive session where you can chat with the AI about your code.
Yes! Use the --prompt flag:
forge --prompt "Explain what this function does" src/utils.js
Or use specialized commands:
forge cmd "list all JavaScript files modified in the last week"
forge commit  # Generate a commit message
forge suggest # Get a command suggestion
Slash commands are special commands you can use during an interactive session:
  • /model - Switch AI models
  • /clear - Clear conversation history
  • /save - Save current conversation
  • /load - Load a conversation
  • /exit - Exit Forge
  • :copy - Copy last assistant message to clipboard
Type /help in an interactive session to see all available commands.
Forge maintains context across your session through conversations:
  • Conversations persist automatically
  • Each project can have multiple conversations
  • Use forge conversation list to view all conversations
  • Resume conversations with forge conversation load <id>
  • Export conversations with forge conversation export <id>
You can limit stored conversations with:
export FORGE_MAX_CONVERSATIONS=100
Restricted mode (forge --restricted) enables enhanced security by limiting shell and file system operations. Use this in untrusted environments or when working with sensitive codebases.
Yes, but with safeguards:
  • Forge will show you commands before executing them
  • You can review and approve each command
  • Some tools have built-in safety measures
  • Use --restricted mode for additional control
Configure command execution behavior in forge.yaml:
max_requests_per_turn: 50
max_tool_failure_per_turn: 3

​
Model Context Protocol (MCP)

Model Context Protocol (MCP) is a standard for connecting AI agents to external tools and services. It allows Forge to integrate with:
  • Web browsers for automation
  • External APIs
  • Custom development tools
  • Third-party services
Learn more in the MCP documentation.
Use the MCP commands:
# Interactive setup
forge mcp add

# JSON-based setup
forge mcp add-json

# List servers
forge mcp list

# Remove a server
forge mcp remove <server-name>
Or manually edit .mcp.json:
{
  "mcpServers": {
    "my-server": {
      "command": "node",
      "args": ["server.js"],
      "env": { "PORT": "3000" }
    }
  }
}

​
Integration

The ZSH plugin provides shell integration features. Set it up with:
forge setup
Or manually add to your .zshrc:
# Load Forge ZSH plugin
source ~/.forge/shell-plugin/forge.zsh
Then reload your shell:
source ~/.zshrc
Yes! While Forge primarily operates in the terminal, you can:
  • Use Forge alongside VS Code
  • Open files in VS Code from Forge
  • Use VS Code’s integrated terminal to run Forge
See the VS Code integration guide for details.
Yes! Forge can be integrated into CI/CD workflows for:
  • Automated code reviews
  • Test generation
  • Documentation updates
  • Code analysis
See the GitHub Actions guide for examples.

​
Troubleshooting

Try these steps:
  1. Check if a tool operation timed out (default: 300s)
  2. Run with verbose logging: forge --verbose
  3. Reduce search limits in environment variables
  4. Adjust timeout: export FORGE_TOOL_TIMEOUT=600
  5. Check your internet connection
  6. Verify the AI provider service status
Enable debug logging:
export FORGE_LOG=forge=debug
forge --verbose
To debug HTTP requests:
export FORGE_DEBUG_REQUESTS=/tmp/forge-debug.json
Check the troubleshooting guide for more solutions.
For development environments, you can:
  1. Add custom root certificates:
export FORGE_HTTP_ROOT_CERT_PATHS=/path/to/cert.pem
  1. (Not recommended) Disable certificate validation:
export FORGE_HTTP_ACCEPT_INVALID_CERTS=true
⚠️ Only disable certificate validation in trusted development environments.
  1. Verify your API key is correct and active
  2. Check the provider’s service status
  3. Ensure your API key has the necessary permissions
  4. Try logging out and back in:
forge provider logout
forge provider login

​
Costs & Pricing

Forge itself is free and open-source. However, you’ll need to pay for:
  • AI provider API usage (pay-as-you-go)
  • Optional: Forge Services subscription for managed AI access
Costs vary by provider and model. Most providers charge per token used.
Control costs by:
  1. Setting request limits in forge.yaml:
max_requests_per_turn: 25  # Lower limit
  1. Using less expensive models:
forge --model claude-3.5-sonnet  # Instead of claude-3.7-sonnet
  1. Limiting conversation history
  2. Using semantic search efficiently
  3. Monitoring usage through your provider’s dashboard
The ZSH theme can display cost information. Configure it with:
export FORGE_CURRENCY_SYMBOL="$"
export FORGE_CURRENCY_CONVERSION_RATE=1.0
Cost tracking depends on your AI provider’s API response including usage data.

​
Contributing

Contributions are welcome! You can:
  • Report bugs on GitHub Issues
  • Submit feature requests
  • Contribute code via pull requests
  • Improve documentation
  • Help others on Discord
See the contributing guide for details.
Yes, contributors are required to sign a Contributor License Agreement (CLA). The CLA assistant will guide you through this process when you submit your first pull request.
Please report security issues privately by emailing the maintainers or using GitHub’s security advisory feature. Do not create public issues for security vulnerabilities.

​
Community & Support

  • Discord: Join our community at discord.gg/kRZBPpkgwq
  • GitHub Issues: Report bugs and request features
  • Documentation: Browse comprehensive guides
  • Discussions: Ask questions on GitHub Discussions
Yes! Join our active Discord community to:
  • Get help from other users
  • Share tips and workflows
  • Provide feedback
  • Stay updated on new features
  • Connect with the development team
Discord

​
Still have questions?

If your question isn’t answered here:

Build docs developers (and LLMs) love

Get started for freeTalk to us