17 New AI Tools Worth Trying in 2026

The AI tool market stopped being interesting when everything was "chat with a document". What changed in 2026 is that tools started finishing work instead of suggesting it. Agents open pull requests, video models render clips with sound, and inbox tools answer mail in your own voice.

This list is not a ranking. It is a set of tools that either launched recently or became genuinely usable this year, grouped by what you would actually hire them for. Every tool here links to its own page on this site, where you can check pricing and jump straight to the official site.

Coding: agents that finish the ticket

Jules — Google's asynchronous coding agent. You point it at a GitHub repository, describe the task, and it clones the repo into a cloud VM, writes a plan, produces a diff and opens a pull request. It is built for the work you keep postponing: bug fixes, dependency bumps, missing tests. Free tier with daily task limits, larger quotas on the Pro and Ultra plans.

OpenAI Codex — OpenAI's coding agent, available as a cloud agent that runs several tasks in parallel and as a CLI you can run in your own terminal. It reads the repository, writes and runs code, and hands back a diff to review instead of silently editing files.

OpenHands — the open-source option. An agent platform you can run locally, in CI, or hosted, that edits files, runs commands, browses the web and opens pull requests. If your code cannot leave your network, this is the realistic starting point.

Continue.dev — an open-source assistant for VS Code, JetBrains and the terminal. The useful part is that you bring your own model and define agents and rules once, then share that configuration with a team through a hub.

Augment Code — aimed at very large codebases. A context engine indexes millions of lines so multi-file refactors and pull-request review do not fall apart once a repository outgrows a single context window.

Bolt.new — a browser app builder from StackBlitz. Describe an app and it scaffolds, runs and deploys it with a live preview, while leaving the generated code editable. Good for prototypes you intend to outgrow.

Trae — ByteDance's AI-native IDE for macOS and Windows, with a builder mode that plans multi-file features and an agent that handles terminal and debugging work.

Video: from five-second clips to usable shots

Google Flow — Google's filmmaking tool built on Veo and Imagen. You generate scenes, control camera moves, keep characters consistent between shots and assemble everything on a timeline. It is closer to a pre-visualization tool than a one-click generator.

Google Veo — the model family behind Flow, also reachable through Gemini and the API. Current versions generate clips with native audio, camera control and reference images.

Kling AI — Kuaishou's platform for text-to-video and image-to-video, with camera control and a motion brush. Consistently strong on motion, and there is an API if you need to batch.

Hailuo AI — MiniMax's generator, known for character motion, with free daily credits that make it cheap to test before committing.

LTX Studio — Lightricks' storyboard-to-video studio. You build the shot list first, then render, which suits ad concepts and client previews better than prompt roulette.

Images: one canvas, many models

Krea AI — a canvas that bundles many current image and video models in one place, with realtime rendering and enhancement up to 21K. Useful precisely because it lets you compare models without signing up for each one.

Magnific — an upscaler that reimagines texture and lighting rather than just sharpening pixels, with sliders to control how far it drifts from your original.

FLUX — Black Forest Labs' image models, available through their playground, an API and most third-party hosts.

Freepik AI — generation and editing layered on top of a stock library, which matters if you need licensing you can show a client.

Remove.bg — still the fastest path from product photo to cutout, with an API and plugins for automated pipelines.

Audio: voices that survive scrutiny

Murf AI — a voiceover studio with 200+ voices across 20+ languages and a timeline for syncing narration to slides or video.

Resemble AI — voice cloning, speech-to-speech and real-time synthesis, used for games, dubbing and contact-center products where you need your own voice, not a stock one.

Adobe Podcast AI — free browser tools, including Enhance Speech, which turns a laptop-mic recording into something usable. Still the cheapest audio upgrade available.

Riverside — records each guest locally so a bad connection does not ruin the audio, then transcribes and cuts the session into clips.

Email, calendar and meetings

Otter.ai — joins Zoom, Teams and Meet, transcribes live, and returns summaries and action items.

Reclaim AI — defends focus time and reschedules tasks and habits around meetings automatically.

Fyxer — an email assistant that drafts replies in your voice and triages the inbox, working over Gmail and Outlook.

Shortwave — an email client where search is natural language and the assistant drafts from your own sent history.

Genspark — an agent workspace that returns a structured research page rather than a chat thread, and can also produce decks and spreadsheets.

Monica — a browser-side assistant that summarizes pages, translates, writes and generates images without leaving the tab you are on.

Agents for teams

Dust — build assistants connected to Slack, Notion, Drive and GitHub, with shared workspaces for a team.

Relevance AI — a no-code builder for teams of agents, with a visual workflow editor and a tool library.

Clay — data enrichment plus AI research agents that run per row, which is how people are building outbound lists now.

How to choose

Pick by failure mode, not by feature list. If the work is code you do not want to write, use an agent that produces a pull request — Jules or OpenAI Codex for hosted work, OpenHands if it must stay inside your network. If the work is video, decide whether you need a single clip (Hailuo AI, Kling AI) or a sequence with continuity (Google Flow, LTX Studio). If the work is email, the question is only whether you want a new client (Shortwave) or a layer on the one you already use (Fyxer).

Most of these tools have a free tier. Test one against a real task this week rather than collecting logins.