Show HN: Open-source Antigravity plugin for Claude Code

The Plugin That Breaks the Terminal Ceiling

A developer has released an open-source plugin for Claude Code that integrates what they call “antigravity” capabilities directly into the command-line interface. The plugin adds slash commands for native image generation, deep research, and code review without requiring the user to leave the terminal environment. While the summary hints at a playful name, the underlying functionality is serious: it extends Claude Code’s reach into multimodal and research-heavy workflows that previously demanded separate tools or browser sessions.

What Actually Happened

The plugin, shared on Hacker News under “Show HN,” is a practical extension for Claude Code, Codex, and similar agentic coding tools. By adding custom slash commands, it allows users to generate images, conduct deep research, and perform structured code reviews—all from within the same terminal session where they write and debug code. The “antigravity” branding appears to be a metaphor for lifting the agent beyond its usual text-and-code constraints. The plugin is open-source, meaning others can inspect, modify, and build upon it.

Why It Matters

This is more than a novelty. For AI practitioners who live in the terminal, the plugin addresses a persistent friction point: context switching. Every time a developer has to leave Claude Code to open a browser for image generation or to run a separate research tool, they lose momentum. The plugin collapses those steps into a single interface, effectively turning Claude Code into a more unified agentic workspace.

The inclusion of native image generation is particularly noteworthy. Most coding agents remain text-only, yet visual outputs—diagrams, UI mockups, architecture sketches—are increasingly part of the development workflow. By enabling this from within the terminal, the plugin hints at a future where agents are not just code generators but full-stack creative partners.

The deep research command also signals a shift. Agents that can autonomously search, synthesize, and return findings without leaving the coding environment reduce the cognitive load on the developer. This is especially valuable for tasks like dependency analysis, API documentation lookup, or competitive research during development.

Implications for AI Practitioners

For developers using Claude Code or Codex, this plugin demonstrates a growing ecosystem of extensibility. The open-source nature means it can be forked, customized, or integrated into team workflows. It also raises the bar for what a terminal-based agent should offer: if a plugin can add image generation and research, why aren’t these features built-in?

Practitioners should note that this approach relies on the underlying model’s ability to handle multimodal inputs and outputs. The plugin is a thin layer—it works because Claude itself can process image generation prompts and research queries. The real innovation is in the UX design: reducing friction by keeping the user in the flow.

However, there are trade-offs. Running image generation or deep research from the terminal may consume significant API tokens or compute resources. Practitioners will need to weigh the convenience against cost, especially for teams on usage-based pricing.

Key Takeaways

• An open-source plugin now adds image generation, deep research, and code review slash commands to Claude Code and similar agents, eliminating the need to leave the terminal for these tasks.
• The plugin reduces context switching, turning Claude Code into a more unified agentic workspace that handles both code and creative/research outputs.
• This signals a growing expectation that terminal-based AI agents should support multimodal and research workflows natively, not just text and code.
• Practitioners should evaluate the token and cost implications of running these extended capabilities, especially in team or production environments.