Forge.ai

Local coding agent. Reads your files, runs shell commands, edits code, talks to your chosen LLM. Runs entirely on your machine — your code never leaves it.

Install & run

npx @theforge-ai/forge --open

That's it. First run downloads the matching binary (~20 MB) into ~/.forgeai/binaries/, then launches the server and opens http://localhost:8787 in your browser.

Subsequent runs skip the download.

Or install globally so you can just type forge-ai:

npm i -g @theforge-ai/forge
forge-ai --open

(The CLI command is forge-ai — not plain forge — to avoid the common collision with Foundry's forge Ethereum dev tool.)

What you'll see

Forge.ai is ready → http://localhost:8787
Open this URL in your browser. Press Ctrl-C to stop.

In the browser:

1. Click the gear icon (bottom-left) → Settings
2. Under Models, paste an API key for any provider (Anthropic / OpenAI / DeepSeek / xAI / etc.)
3. Top-left folder button → point at the project you want the agent to work on
4. Type a prompt, hit Enter

Default permission mode is prompt — every tool call surfaces a confirmation. Change it from the MODE badge in the input area once you trust the setup.

Flags

forge-ai --help              # show all flags
forge-ai --open              # auto-open browser
forge-ai --port 0            # auto-pick a free port
forge-ai --port 4000         # bind to a specific port
forge-ai --host 0.0.0.0      # expose on LAN (NOT recommended)

State

• ~/.claw/state.json — config, providers, MCPs (persists across runs)
• ~/.claw/libraries/ — RAG library sqlite stores
• ~/.claw/skills/ — user-level skills
• ~/.forgeai/binaries/v<version>/ — downloaded binary cache

Updating

npx @theforge-ai/forge@latest --open

To clear the cached binary for a version:

rm -rf ~/.forgeai/binaries/v0.1.0

Supported platforms

• macOS arm64 (Apple silicon)
• macOS x86_64 (Intel)
• Linux x86_64

Other targets: open an issue.

License

Apache-2.0