DEV Community

Cover image for ⚡9 AI Tools you cannot afford to miss in 2025! 🚀🔥
Shrijal Acharya for Composio

Posted on • Originally published at composio.dev

⚡9 AI Tools you cannot afford to miss in 2025! 🚀🔥

This article lists my top 9 AI tools I believe every developer or someone interested in AI should use to get the most out of AI in 2025. ✅

These tools are aimed at solving real problems, offering real value, and making life easier. I've compiled a list of such tools that I've tried myself and use frequently. Some of them are well-known, while others are underrated.

If you build software or are interested in AI, chances are a couple of these will be in your daily toolkit. 🤫

1. Rube - Chat with anything in your workflow

ℹ️ MCP server that connects your AI tools to 500+ apps.

Rube MCP

You probably already have an idea of what an MCP is, right? If not, it stands for Model Context Protocol and think of it as a bridge between AI models and external tools, which can provide it with data and the ability to take action on them.

An MCP server is this smart adapter for these tools, which the MCP clients can connect to (Claude, Cursor, etc.).

MCP Working

Rube's MCP server acts like a remotely hosted MCP server that includes all these applications you can use (Slack, Gmail, Facebook, etc.). You name it, and it's all there. You get access to over 500 apps in your chat.

Visit the marketplace to see a list of all the apps you can use with Rube.

You can get started with Rube by installing it on your platform of choice.

Rube Installation

Or, simply sign up in the Rube web app, set up the connection for your application of choice, and test it directly in your browser.

Check out this quick demo to get an idea of what Rube can do. 👇


2. Veo 3 - AI Video Generation with Native Audio

ℹ️ Google DeepMind’s latest model for generating short, high-quality videos from text prompts, with audio support.

Veo 3

Veo 3 gives you short, polished clips you can actually ship (it’s so real). You describe the shot. It handles motion, lighting, and even the audio so the scene doesn’t feel silent or awkward. Outputs are short, clean, and great for shorts, or to turn your prompt into a little cinematic clip.

What you can do right now:

  • Generate 8-second clips at 720p or 1080p from text, directly from Gemini Apps or the API.
  • Get native sound effects, ambient noise, and even simple dialogue in the same pass.
  • Try the lighter Veo 3 Fast inside YouTube Shorts for quick prompts on your phone. It runs at lower latency and 480p, so you can build on your ideas faster. Everything is labeled with SynthID (which is a way Google uses to watermark and identify AI-generated content).

Quick Start:

  • In Gemini: open the video tab, type a short description, and render. (Make sure to have the Google AI Pro Plan)
  • For devs: call Veo 3 through the Gemini API or Vertex AI to generate clips programmatically. You can find more details here: Veo 3 API.

Check out this video on "Sailer and the Sea," generated by Veo 3 (note the quality of the video and the audio, insane!). 👇


3. Kombai - AI Agent built for Frontend

ℹ️ Kombai is a front-end AI agent that for real world Frontend tasks.

Kombai - AI Agent for frontend

Kombai is an AI agent built specifically for frontend work. It takes input from Figma, text, images, or existing code, understands your stack, and generates production-ready UI with solid structure and styles. It’s tuned for real frontend tasks, not just generating similar UI.

You can install it right in your editor. It works in VSCode, Cursor, Windsurf, and Trae. Launch it from the extension marketplace, and you're set.

With Kombai, you can:

  • Convert Figma designs into code (React, HTML, CSS, etc.) using your preferred component library that suits your project.
  • Use a frontend-aware engine that knows 30+ libraries like Next.js, MUI, Chakra UI, and more.
  • Iterate in-editor, keep your conventions, and ship faster with higher fidelity to the design.

Visit the docs to get started and see the setup for your editor of choice.

You can get going in a minute:

  • Install the extension for your editor
  • Sign in, connect your project
  • Paste a Figma link or describe the component you want
  • Review the code and commit

If your day is mostly frontend, this is a must-have.

Check out this quick demo to get an idea of what Kombai can do. 👇


4. ElevenLabs - AI Voice Synthesis

ℹ️ The most realisitic and high-quality voice AI platform.

ElevenLabs

If you've ever wanted something to say what you write in a voice that sounds human, expressive, and maybe even like you, ElevenLabs does exactly that. It's not just text-to-speech (although it does that extremely well); it lets you clone voices, create audiobooks, build music, and all kinds of creators are using it more than ever.

Creators using ElevenLabs

With ElevenLabs, you can build workflows that:

  • Clone a voice from just minutes of audio, then use that voice across podcasts, videos, or narration.
  • Turn written content into natural speech in many languages and accents.
  • Add emotion, tone shifts, or pauses inside speech (whisper, excited, sad, etc.) so narration feels more human.

If you've played Chess on chess.com, you've probably already experienced ElevenLabs. All the voices you hear are from ElevenLabs; notice how real that sounds?

For any projects that require me to add text-to-speech or work with any form of audio creation, I usually use ElevenLabs. They also have a great free tier. Check out their pricing to learn more.


5. Cursor - The AI Code Editor

ℹ️ Edit your code using natural language to build software faster.

Cursor

Cursor is an AI-first code editor that fully understands your repository. You tell it what you want in natural language, and it updates files, refactors safely, answers questions, and even runs commands in your terminal with your approval, all with full project context.

ℹ️ The whole vibe coding thing really started to get attention with the release of Cursor.

It is a fork of VSCode, so everything will feel familiar to most of you. You can bring your own extensions, themes, and keybindings.

The platform has also recently raised $60 million in funding, which underscores the significance of this IDE.

With Cursor, you can:

  • Edit code in plain English, apply multi-line changes, and accept them with Tab.
  • Ask about parts of your codebase and get fixes that reference the right files.
  • Use Agent to execute terminal commands, loop on errors, and iterate toward a working result.
  • Plug in MCP tools to reach outside your repo, for example, pulling design context from Figma or querying services via MCP servers (e.g., Rube or any other local MCP servers).

There is a lot that you can do with Cursor; this is just to give an idea.

💡 If you're someone who prefers a GUI and wants AI directly in your code editor, then Cursor is the best choice.


6. OpenCode - AI Coding Agent for the Terminal

ℹ️ OpenCode is an AI coding agent that works in the terminal.

OpenCode

If Cursor brings AI directly into your editor, OpenCode does something similar but for the terminal. It’s built for people who live in their terminal and don’t want to leave it, especially developers who love working in Neovim or prefer a CLI-first workflow (like me 😉).

With OpenCode, you can:

  • Run it as your AI coding agent right inside the terminal.
  • Get Cursor-like functionality without leaving your shell.
  • Choose the AI model you want to work with (Claude, GPT, Gemini, etc.).

Additionally, consider OpenCode as a single replacement for multiple CLI coding agents, such as Gemini CLI, Claude Code, or OpenAI Codex. So instead of juggling many different CLI tools, you just have one AI coding agent in your terminal that does it all.

You can get started with OpenCode by running the following command:

# YOLO
curl -fsSL https://opencode.ai/install | bash

# Package managers
npm i -g opencode-ai@latest        # or bun/pnpm/yarn
brew install sst/tap/opencode      # macOS and Linux
paru -S opencode-bin               # Arch Linux
Enter fullscreen mode Exit fullscreen mode

Visit their documentation for more information.


7. Poke - A contact that acts on your behalf

ℹ️ Poke is a contact that acts on your behalf with whom you can communicate.

Poke

Think of Poke as a contact in your phone that you can actually delegate to. You text it in iMessage, WhatsApp, or SMS. It connects to your email, calendar, and files, then turns the chats into real actions. No new app to learn. Just message it, and it gets things done.

🤔 Why is this interesting?

Poke launched publicly in September 2025 and raised $15M led by General Catalyst, valuing the company at around $100M. The bet is simple. Most people do not want another app. They want help inside the apps they already use all day.

You can try it directly in your messages. Start a thread with Poke, connect your accounts, and give it a small task. It will propose actions in short bubbles you can approve. If you like the result, hand it more.

So, the workflow is simple: Text it -> Approve it -> Done.

Here's a quick film to learn more about Poke. 👇


8. NotebookLM - AI Research Tool & Thinking Partner

ℹ️ Upload your sources, ask questions, and get cited answers, even an audio summary you can listen to.

NotebookLM

NotebookLM is like a research buddy. You load it up with your stuff, and it learns only from that. It then answers with citations, connects ideas, and can generate summaries you can read or listen to. It runs on Google’s Gemini models and is built for deep dives, not your usual ChatGPT-styled messages.

With NotebookLM, you can:

  • Add sources from Google Docs, Slides, PDFs, text, web URLs, YouTube, and audio files. Up to 50 sources, with support for large files.
  • Get an Audio Overview. Two AI hosts guide you through your materials, providing a concise overview like a short podcast that you can skim by listening.
  • Ask questions and get answers based on your sources, with inline citations you can click.
  • And there's so much more...

Quick Start

  • Create a notebook at notebooklm.google
  • Add a few core sources first
  • Ask a specific question, then drill down with follow-ups
  • Generate a study guide or Audio Overview to review faster

Here's a video introduction to NotebookLM. 👇


9. Manus - General AI Agent

ℹ️ Give it a goal. It plans, executes, and circles back with results.

Manus

Manus pitches itself as a general AI agent that turns thoughts into actions. You hand it a goal. It figures out the steps, performs the work, and reports back. Research, plan, draft, deploy, even spin up small apps.

With Manus, you can:

  • Set a high-level goal and let it run multi-step tasks end-to-end.
  • Offload research and get a concise report with sources.
  • Build a full website from a short brief and ship it fast.
  • Translate a PDF while keeping the original formatting.

This is kind of an all-in-one agent. One place to build sites, translate documents, draft outlines, analyze data, and a bunch more. Browse the Playbooks to see what’s possible and launch with one click.

Quick start:

  • Create an account on manus.im
  • Open a playbook or write your own goal
  • Review the plan, approve steps, and let it run.

Here's a quick video introduction to Manus. 👇


If you think of any other handy AI tools that I haven't covered in this article, please share them in the comments section below. 👇🏻

That concludes this article. Thank you so much for reading! 🫡

Ryan Gosling Bye Bye

Top comments (11)

Collapse
 
aayyusshh_69 profile image
Aayush Pokharel

If you're a developer and want some other best AI picks for coding. Here's some of choices of mine:

  1. Cline
  2. Tabnine
  3. SuperMaven
  4. Instead of OpenCode, I use Claude Code specifically. Gemini CLI helps
  5. Aider for git stuffs.
Collapse
 
shricodev profile image
Shrijal Acharya Composio

These are great tools. Thanks, buddy!

Collapse
 
cyber8080 profile image
Cyber Safety Zone

Nice list! A few thoughts that came to mind after reading:

  • I really like how this article balances between "well-known AI tools"(like ElevenLabs, Cursor) and some "underrated ones" (e.g. Rube, OpenCode). Gives both reliability and discovery.
  • Kombai sounds especially useful for frontend devs — turning Figma or descriptive ideas straight into production UI is a huge time saver.
  • The “general AI agent” angle (Manus) is exciting — having one tool that can plan, execute, and loop on tasks could reduce a lot of context switching.
  • One thing to keep in mind is 'integration & workflow fit".A tool is only as good as how seamlessly it fits into what you already use (editor, CI/CD, etc.).
  • I’d love to hear more from folks who’ve actually adopted one of these for weeks — what’s been the biggest friction or surprising benefit?

Thanks for sharing this — I’m going to try a couple of the tools mentioned and see how they change my dev routine.

Collapse
 
shricodev profile image
Shrijal Acharya Composio

Glad you liked it! Do give these a shot and let me know how it goes.

Collapse
 
shricodev profile image
Shrijal Acharya Composio

If you think of any other handy AI tools that I haven't covered in this article, please share them here.

Collapse
 
nabin_bd01 profile image
Nabin Bhardwaj

Thanks for sharing. I'd love to try out Kombai for frontend

Collapse
 
shricodev profile image
Shrijal Acharya Composio

Sure, give it a shot. Let me know how it turns out. I've tested some frontend stuff with Figma MCP as well, which you can find here: Reddit: It's been nice knowing you frontend devs

But Kombai got me the best result, and this is not even close. Claude + Figma MCP didn't stand a chance; you can imagine how good the result turns out with Kombai. I'll most likely share some nice demos in the near future.

Collapse
 
sonu_goswami profile image
Sonu Goswami

This is an incredible roundup!

Collapse
 
shricodev profile image
Shrijal Acharya Composio

Glad you liked it!

Some comments may only be visible to logged-in visitors. Sign in to view all comments.