Forge Code

What is Forge Code?

Forge Code is an AI coding agent designed for people who want the power of tools like Claude Code without being locked into one model or one vendor. You run it from the command line, point it at your repository, and it reads your files, plans the work, then makes the edits for you. An IDE extension brings the same agent into your editor for a more visual workflow.

The project is open source and built around a bring your own model philosophy. You add API keys for the providers you already use, or point it at a local model, and Forge handles the orchestration: reading context, writing files, running commands and showing you a diff of everything it changed. That makes it popular with developers who want agentic coding on their own terms.

It is part of a wave of terminal-first coding agents that appeared in 2025 as a response to the popularity of Claude Code and OpenAI Codex, and it aims to be lighter and more hackable than the big commercial tools.

Key Features

• Terminal-based agent that reads, plans and edits code across a whole project
• IDE extension for working alongside your normal editor workflow
• Multi-model support, including Claude, GPT, Gemini and local models via API keys
• Model Context Protocol (MCP) support for connecting external tools
• Clear diffs and review prompts before changes are applied
• Configurable through plain config files you can commit to your repo
• Lightweight install that runs on modest hardware

How to get started

Getting started with Forge Code is quick. Head to the official website, grab the installer for your platform, and add at least one model API key in the configuration step. Then open a terminal inside your project directory and give the agent a task.

Start small: ask it to fix a bug or add a test, review the diff it proposes, and approve. Once you are comfortable, try the IDE extension and let Forge take on larger refactors. The project documentation covers setup, model configuration and the extension.

• Install the Forge CLI from forgecode.ai
• Configure your model provider with an API key
• Open a terminal in your project folder
• Describe a task, for example fix a failing test
• Review the proposed diff and approve the changes
• Install the IDE extension for a visual workflow

Use cases

• Refactoring and renaming across large codebases
• Fixing bugs and writing regression tests
• Scaffolding new features and boilerplate
• Exploring and explaining unfamiliar code
• Generating documentation from existing code
• Repetitive edits across many files

Pricing and licensing

Forge Code is open source and free to use. You pay only for the model usage behind your own API keys, which makes it a very affordable option if you already have accounts with Claude, OpenAI, Google or a local inference setup.

Because you bring your own keys, there is no subscription and no per-seat pricing. Check the official website and the project repository for the latest details on supported models and any hosted offerings.

Pros and cons

The biggest wins are freedom and price. Open source means you can inspect the code, contribute, and run it without vendor lock-in, and the bring your own key model keeps costs predictable. The terminal-first design also suits people who already live in the command line.

The trade-offs are setup effort and support. You have to manage your own API keys and you miss the polished, all-in-one experience of a commercial service. The ecosystem is younger too, so fewer tutorials exist compared with the big-name agents.

Alternatives

• Claude Code: Anthropic's agentic coding tool, included with Claude subscriptions
• OpenAI Codex: agentic coding from the makers of ChatGPT
• Aider: an older open source terminal pair programmer
• Cursor: an AI-native code editor with built-in agents
• Google Gemini CLI: free and open source agentic coding from Google