Head-to-Head: Gemini CLI vs Create Anything in Agentic AI

I've been watching the agentic development world buzz with excitement lately. There are so many new entrants and big tech giants all fighting for dominance and user adoption in this space. Even the developer CLI segment, which I always considered a bit of a safe haven, has seen a ton of AI action, with recent launches like Gemini CLI in late-June 2025 and Claude Code in May 2025. And, of course, OpenAI has Codex CLI.

But then, a new player caught my eye: (Create) Anything. This is a web-based AI app builder that promises to let users "Create powerful apps & websites by chatting with AI". While it does not provide a terminal CLI. my initial trial with it, although short-lived, really impressed me. Naturally, I had to see how the Gemini CLI Coding Agent measured up against Anything.

My Challenge: A Portfolio Website for a Product Manager

To truly assess what these agents could do, I gave each of them a moderately complex task: create a portfolio website for a software product manager. This task required some structured thinking, but at least with my initial prompt, it didn't involve databases or authentication, it was purely a front-end project from scratch.

I gave both agents the exact same prompt:

Create a portfolio website for me, a software product manager. The website should have:

• a home page with a large headshot image, a brief paragraph about myself, a shortcut to my LinkedIn profile.

• a projects landing page which lists my projects, in a grid format like products on an e-commerce site. Each project will have a project image, a title, and a short 50-word description.

• Clicking on the project will open a project detail page with more text and images or screenshots or links.

• There should also be a landing page for blogs. Each blog should have its own page that would support typical blog content - rich text, images, gifs, embedded videos, etc."

Since I don't have the technical depth to analyze code quality or security vulnerabilities, my comparison was purely functional – I focused on the "first-cut user experience".

How the Coding Agents Fared

Let me break down my observations page by page.

Note that the images and content on the portfolio website are dummy - the projects and blogs are sample data created by the AI agents and the images have been taken from unsplash.

Home or Landing Page

(create) Anything provided an impressive start. Anything created two applications – one for web and another mobile-friendly one, which I didn't even specify in the prompt. The web version appeared to utilize a React framework with Vite and Tailwind CSS, while the mobile version seemed to be based on React Native. Its landing page in the web version looked way more polished than Gemini CLI's and included a section for summary metrics like "#years of experience" and "#products". The LinkedIn call to action button is also more polished.

Gemini CLI created a single application based on React framework with Bootstrap CSS. It did an "okayish job". The UI wasn't very polished, and the nav-bar resembled a typical basic React nav-bar. The page also did not make good use of vertical space with a lot of empty space below the image and text content.

Projects Page

Anything’s project landing page was equally impressive. It featured a Page title with a short intro, and the project cards had a consistent look and feel with uniform image sizes. It also included tags for categorizing projects (e.g., Mobile App, AI/ML). What truly showed "some real thought" was the animation on hover.

Gemini CLI: Here, I observed the gap really start to widen, as its version was just shoddy with inconsistent card image sizes.

The difference became even more apparent when I compared the project details page. Anything’s version was superb and so thoughtful given the vague prompt. Anything added a title, tag, duration, team size, and key impact, followed by a header image. It then logically structured the page with sections for Overview, Challenge, Solution, Screenshots, and Results and Impact. It even thoughtfully included a section with fictional project links for app stores.

Anything Project Page

Gemini CLI’s version of project page was unprofessional and barely passable featuring a huge header image immediately followed by a short project description, completely lacking any sense of sections or formatting.

Blogs Page

My observations here were quite similar to the project pages.

• Anything: Its initial version was super thoughtful presenting a featured blog section followed by an "All Articles" section. These article cards were consistent, showing categories/tags, published date, and estimated read time.

• Gemini CLI: This initial version would require a lot more iterations to refine.

Here’s the Blogs Landing Page from the Anything project:

Here’s the version created by Gemini

Finally the sample blog pages generated by Anything and Gemini respectively:

Anything

Gemini:

My Verdict: Polish vs. Potential

So, what's my conclusion from this initial trial?

Anything truly does a great job despite a prompt with not too much detail. It created a mobile-friendly version, and each page was thoughtful and very polished. I believe this first-cut could easily be published with minor edits. My initial trial, though described as "short-lived," was remarkably impressive.

Gemini CLI's first cut was just okay. It would clearly require a much more comprehensive prompt upfront and likely multiple iterations to fine-tune the user experience.

However, there's a crucial point where Gemini CLI would likely stand out, something I didn't cover in detail in this comparison: its versatility. With support for MCPs (Model Context Protocol), if I were to provide Gemini CLI with screenshots or Figma designs, I believe its first-cut output could be significantly better. It also boasts a good depth of tools, including web-search.

Another huge plus for Gemini CLI is the generosity of its free tier. It offers 60 model requests per minute, 1000 per day, and a massive 1 million token context window – a capacity that could serve many complex daily requirements. In contrast, Anything's free tier credits are, understandably, very limited, and mine were exhausted after just a few prompts. I'm also not sure regarding Anything's current MCP support.

In summary, for a quick, polished, and thoughtful first-draft app generated from a vague prompt, Anything is a clear winner in my experience. But for developers who want more control, can provide visual inputs, and need a powerful, versatile tool with a generous free tier, Gemini CLI offers compelling potential.