ScroogeFrog AI helps advertisers find out where their ad budget is going to waste. It audits website traffic and Google Ads campaigns to uncover suspicious clicks, bot activity, and underperforming spend, then turns complex data into clear recommendations on what to fix. For agencies and larger advertisers, it also offers custom AI fraud detection models trained on their own traffic data. Less guesswork, more control over every advertising dollar.

Nod does the work around your work, in your apps and on real websites, using a private computer of its own on your Mac or PC. It takes orders only from you, asks before anything you didn't ask for, and runs on the AI provider you choose. Free, Apache 2.0.

CALL-E lets your AI assistant make phone calls for you. Ask ChatGPT or Claude to book a table, get a quote, move an appointment or check stock. CALL-E asks for anything missing, shows you the plan, and dials a real US or Canadian number after you confirm. It gets through phone menus, hold and voicemail, then brings back a summary, the transcript and what happened. Builders can add it to Claude Code, Codex, Cursor or OpenClaw with one prompt. New users get 100 free credits.

Scribe v3 is ElevenLabs’ most accurate speech-to-text model, for developers building on calls, meetings, interviews and media. It makes 27% fewer word errors than Scribe v2 across 90+ languages. Unlike transcription that breaks on noise, accents and crosstalk, it labels up to 32 speakers, tags laughter, music and emotion, boosts up to 1,000 of your own terms, detects 56 entity types and handles files up to 15 hours. For live use, Scribe v3 Live runs under 150ms.

Skillbox keeps your AI skills in one library and syncs them to Claude Code, Cursor and Codex. Follow GitHub sources, review diffs, and scan each version for prompt injection. Trim shortens noisy command output so sessions last longer.

MiniMax H3-Image is an open-weight model for native 16MP image generation and editing. It uses just six diffusion steps, handles text-to-image and editing in one checkpoint, and supports up to 16 reference images.

Most people record with one camera, because more angles feel too hard. Shot.video turns your laptop and phones into a multi-camera studio. Scan a QR code and a phone becomes another camera, with no app to install. Press record once. Every angle records in sync, then AI makes your first edit. It cuts to the camera you're looking at and adds zooms, punch-ins and optional B-roll. Change anything you like. Free to start. No card needed.

You send a deck or doc. They ask their AI to read it. The AI sees what's on the page, but not the thinking behind it: why you made those choices, what you left out, or what's still open. Relay shares that context with your document, so they and their AI can understand it and ask questions. You choose what to share. Decisions that need your input, like a discount or a date, come back to you. Try a sample relay. No sign-up needed.

Discover Reddit Reviews, a new experience that organizes product conversations and highlights the most important takeaways from across our communities.

Seedream 5.5 is ByteDance's latest AI image model, built for creators who want to keep refining their work without starting over. Make up to 100 rounds of targeted edits while preserving untouched areas at the pixel level. Generate reusable assets with native transparent backgrounds, create more realistic images with natural lighting and textures, and render detailed text more accurately.

Muse Ember is Meta's smallest multimodal reasoning model, built to handle the repetitive work behind AI agents. From tool calls and coding subtasks to document extraction and routing, Ember delivers efficient execution at scale. It supports text, images, audio, and video, with a 1M-token context window and pricing starting at $0.10 per million input tokens. Pair it with larger models for planning, and let Ember handle the volume. Available through Meta Model API and Muse Code.

For people who save videos, posts, articles and documents across apps and can't find them when they need them, Valin is the private memory app that understands what you save, down to what's said in a video, so you can find information you don't even remember saving. Unlike other AI bookmarking apps, Valin's AI runs entirely on your device, so only you can see what you save. No account, no cloud, no AI subscription.

Fluency is hard, being understood is easier. Kippy is your AI speaking partner for language learning. Not your usual AI sycophant tutor but rather personalities that care just a little more about themselves than you, while they teach you a new language.

Add Stiro to your SaaS with one script tag. When a user gets stuck or asks for something, it does the steps with them in your real product while they watch. Anything that sends, spends or deletes goes back to them. Built to cut how-do-I tickets.

Feeling off but can’t say why? Type it out, or talk it through by voice. Xolace shows you what’s underneath, then hands you steps that fit: a chat with a trained peer supporter, help writing to someone you trust, or guided audio. Reflecting is free.

StableBar is a free, open-source menu bar for macOS written in Swift. Set it up in a Settings window or one JSON file instead of shell scripts. It shows AeroSpace, yabai or OmniWM workspaces, widgets with popups and seven themes, and you edit it on the bar.

Most AI agents can’t call. Those that can often use unfamiliar numbers that get blocked or flagged as spam. Even when someone picks up, a successful voice conversation needs different context handling than a text chat. Mitra MCP lets your agent call from your personal number, with a calling experience built for live conversations and the ability to bring you in when needed.

Connect your email and AI finds every subscription: one list, your monthly total, alerts before renewals and trial ends, cancel steps for 200+ services. Outlook in one click, Gmail and any inbox by forwarding. Never asks for bank logins. Free for 5.

Earlybird is a 29M-parameter, 17 MB on-device speech-to-text model that predicts each word correctly about 90 ms before the speaker finishes it, with a confidence on every word. Built for voice agents that need to act sooner. Open weights and code.

Take a photo of the loose bricks on your table. Mister Blocks (powered by Claude) looks at what you actually have and suggests a few things to build from them, each with 3D step-by-step instructions you can turn, zoom and print. Don't like it? Ask the chat for changes. No pile handy? Describe a build in words and get one, from a robot dog to a voxel sculpture of a sports car. Every build is stud-checked. Free guides work offline.