Getting started
Install Punchcutter in the tool you already use. Then ask for a font.
Every route needs Python 3.10 or newer. Punchcutter installs its own font libraries; nothing else is required.
Claude Code
As a plugin, which adds the skill and the /punchcutter command:
/plugin marketplace add byrishi/punchcutter
/plugin install punchcutter@punchcutter
As a plain skill:
npx skills add byrishi/punchcutter
By hand:
git clone https://github.com/byrishi/punchcutter
mkdir -p ~/.claude/skills && cp -r punchcutter/skills/punchcutter ~/.claude/skills/
pip install -r ~/.claude/skills/punchcutter/scripts/requirements.txt
Then type /punchcutter, or just say "make me a font".
Claude app (web, desktop, mobile)
- Download punchcutter.skill.
- In Claude, open Settings, then Skills, and upload the file. Code execution must be turned on.
- Start a chat: "make me a font".
Claude installs the libraries on first use. When the font is done, the Studio opens right inside the chat.
Claude Desktop, Cursor, Windsurf, VS Code
These connect through MCP. Install the package with the MCP extra:
pip install "punchcutter[mcp] @ git+https://github.com/byrishi/punchcutter"
Add it to the client's MCP config (claude_desktop_config.json, .cursor/mcp.json, .vscode/mcp.json and so on):
{
"mcpServers": {
"punchcutter": { "command": "punchcutter", "args": ["mcp"] }
}
}
Claude Code can add the same server in one line:
claude mcp add punchcutter -- punchcutter mcp
See Claude, MCP and other agents for the tool list.
ChatGPT
ChatGPT connects to remote MCP servers in developer mode (paid plans, on the web). Run Punchcutter as a web server and give ChatGPT its address.
Start the server and open a tunnel to it:
pip install "punchcutter[mcp] @ git+https://github.com/byrishi/punchcutter"
cloudflared tunnel --url http://localhost:8000 # or: ngrok http 8000
punchcutter mcp --http --public-url https://YOUR-TUNNEL-ADDRESS
To host it permanently instead, deploy the repository's Dockerfile to Render, Railway or Fly and set PUNCHCUTTER_PUBLIC_URL to the app's address.
Then connect ChatGPT:
- Open Settings, then Apps & Connectors, then Advanced, and turn on Developer mode.
- Choose Create, paste
https://YOUR-TUNNEL-ADDRESS/mcp, and set authentication to none. - In a chat, pick Punchcutter from the Developer mode tools and say "make me a font".
ChatGPT gets the proof as an image and the Studio and font zip as download links. For the photo routes, give it a link to the photo. Anyone with the address can use the server, so keep the tunnel address to yourself.
Codex
As a skill, from your terminal:
npx skills add byrishi/punchcutter --agent codex
pip install -r ~/.codex/skills/punchcutter/scripts/requirements.txt
Or inside Codex: $skill-installer install https://github.com/byrishi/punchcutter/tree/main/skills/punchcutter. Restart Codex, then type $punchcutter.
As an MCP server:
codex mcp add punchcutter -- punchcutter mcp
or in ~/.codex/config.toml:
[mcp_servers.punchcutter]
command = "punchcutter"
args = ["mcp"]
Gemini CLI, Copilot, Aider
Clone the repository and open your agent in it. The AGENTS.md file tells it how to run the foundry.
git clone https://github.com/byrishi/punchcutter && cd punchcutter
pip install -r skills/punchcutter/scripts/requirements.txt
Then ask: "Follow AGENTS.md and make me a rounded friendly font called Pomelo." The agent must be able to view images, because it checks its own proofs.
No AI at all
The engine is a normal command line tool.
pip install "git+https://github.com/byrishi/punchcutter"
punchcutter new "My Sans" --preset humanist --weights 300,400,700
punchcutter build MySans.json
The CLI reference lists every command.
Platform notes
| System | Note |
|---|---|
| macOS | Proof images and sheets need Cairo: brew install cairo |
| Windows | Use Python from python.org; everything installs with pip |
| Linux | Install libcairo2 from your package manager if proofs fail |
| Claude app | Code execution must be on in Settings |
Your first font
> a calm grotesque for a fintech app, three weights, call it Ledger
Punchcutter answers with its stages, shows you a proof, revises it, and ends with the Studio, the font files and a short list of next moves. That is the whole loop. Next: the six routes, handwriting and photos, and using your font.