· Open Source· GitHub

nitrogen

Carbon.now.sh for AI coding. Compose faux Claude Code, Codex CLI, and Gemini CLI sessions and export crisp, social-ready PNGs — all in your browser, zero backend.

nitrogen is the fastest way to turn a coding-agent session into a polished, shareable image. Pick a layout, add blocks, choose an agent theme, and export a 2× PNG ready for X, LinkedIn, or documentation.

It renders in the style of the leading AI coding agents — Claude Code, Codex CLI, and Gemini CLI — with matching colors, prompt markers, tool labels, and status bars. The exported image looks real enough to post, without requiring any terminal or editor.

nitrogen block editor with live terminal preview
The nitrogen editor — blocks on the left, live preview on the right

Per-agent themes

Each terminal window can be styled as Claude Code (coral dot, Bash/Update labels), Codex CLI (teal dot, apply_patch), or Gemini CLI (blue sparkle, Shell/ReadFile). Themes cover everything: the prompt marker, the permission-mode status bar, the tool-call formatting, and even the corner colors.

Claude Code and Gemini CLI windows rendered side by side
Claude Code vs Gemini CLI, side by side

Compose with typed blocks

Build a session from a structured block editor: user prompts, assistant responses (markdown), Bash commands, Edit diffs, and Read tool calls. Add, reorder, and delete blocks until the conversation looks right. Two windows can be placed side-by-side or stacked, each with its own agent and settings.

Permission modes & framing

Set the status bar to normal, accept edits, plan mode, or bypass permissions — rendered per agent. Then choose a gradient, solid, or transparent backdrop, adjust padding, and pick an aspect ratio: auto, 16:9, square, X/Twitter, or LinkedIn.

Codex CLI session exported by nitrogen on a slate backdrop
Codex CLI export on a slate gradient backdrop

Zero backend, 2× PNG export

Everything runs in the browser. State is persisted to localStorage, and export uses html-to-image to capture the preview at 2× resolution — so the PNG is the exact composition shown on screen, with JetBrains Mono embedded.

Try it at nitrogen.cite-met.dev, or get the source and a bundled agent skill on GitHub → choyiny/nitrogen.