← All transcripts

Top AI Agent Projects : MiniMax Design, Shape, Origin, Viktor & Cherry Blossom Transcript, AI Summary & Key Points

ManuAGI - AutoGPT Tutorials · 3 hours ago · Science & Technology · 20:16 · EN-US

AI Summary

This weekly update covers 20 trending AI-agent projects across creative production, software development, testing, observability, team communication, productivity, scientific research, security, community management, electronics design, sales, and workflow automation. The projects emphasize autonomous execution, consolidated workspaces, local processing, privacy, governance, and integrations with existing tools.

Key Points

  • MiniMax Design runs a local AI crew for creative production, with agents working in parallel on copy, images, video, and voice-over.
  • Aloud converts spoken, screen-recorded feedback into structured plans and tasks for coding agents.
  • Checksum AI generates, runs, and heals Playwright test suites alongside CI/CD using end-to-end, CI, and API agents.
  • Shape combines design, code, Git, and AI chat in one desktop workspace with visual editing and live previews.
  • Bitdrift provides real-time, agentic observability for mobile apps using local telemetry, selective uploads, and privacy-conscious replays.
  • Zoho Cliq 7.0 combines messaging, meetings, automation, AI, webhooks, MCP support, and model selection in one workspace.
  • Wispr Flow turns speech into formatted, auto-edited text across apps and supports more than 100 languages.
  • Fin is an Intercom AI customer agent for support, sales, and e-commerce that can reason through problems, update accounts, and handle payments and refunds.

🔒 12 more in the full analysis

AI in practice

Used for

What
Coordinate copy, image, video, and voice-over agents from one command center.
What
Convert spoken and visual feedback into clear, executable tasks for coding agents.
What
Generate, execute, and repair test suites alongside CI/CD.

🔒 51 more in the full analysis

Agents

  • MiniMax Design — Coordinate creative production across copy, images, video, and voice-over. 2 held 00:59
  • Checksum AI — Generate, execute, and heal automated test suites. 2 held 02:56
  • Fin — Resolve customer support, sales, and e-commerce conversations. 2 held 07:31
  • Cursor agents in Origin — Work on code inside hosted repositories. 2 held 08:28
  • Cluing hosted AI agents — Use shared organizational context to conduct research and build or publish outputs. 2 held 09:26
  • EYE — Manage everyday personal tasks using the user's life data. 2 held 10:25
  • Viktor — Complete delegated business work from research through delivered reports, dashboards, code, and campaigns. 2 held 12:18
  • Vois 2.0 — Run a local voice-production studio from text or through an external AI agent. 2 held 13:13
  • Ressearch AI agents — Execute an end-to-end scientific research workflow. 2 held 14:18
  • Tines agents — Execute sensitive organizational workflows such as alert triage, deployments, approvals, and hiring. 2 held 15:25
  • Antibot — Manage and energize a Discord community. 2 held 16:18
  • Cherry Blossom — Design and validate printed circuit boards from natural-language requirements. 2 held 17:17
  • Clara — Engage, qualify, and convert website visitors into sales meetings. 2 held 18:23
  • Taku AI — Assemble and package AI applications, agents, and skills into reusable workflows. 2 held 19:23

Links mentioned

🔒 Full analysis locked

Unlock more videos and the full analysis

Buy credits to process more videos. Each run includes the full analysis, not just the summary — and you get access to the locked analysis across the library.

Inquire for details →

From this video

31 products

Aloud Antibot Bitdrift bitdrift.ai Checksum AI Cherry Blossom Clara Clara AI SDR Cluing Cluing Cluing Expert Chase Expert Chase Expert Chase Fin MiniMax Design OmniVibe Origin by Cursor Origin Code Hosting Ressearch AI Ressearch AI Shape Taku AI Tines Tines Viktor Vois Vois Wispr Flow Zoho Cliq Zoho Cliq 7.0

Transcript

Searchable transcript of Top AI Agent Projects : MiniMax Design, Shape, Origin, Viktor & Cherry Blossom — ManuAGI - AutoGPT Tutorials (20:16). Search for a phrase, then click its timestamp to jump straight to that moment in the video.

Captions sourced from the original video on YouTube, published by ManuAGI - AutoGPT Tutorials. The video, its captions and all related intellectual property remain the property of their respective owners; AINotes claims no ownership. Provided for research, accessibility and search — see the Transcript Notice and Copyright Policy.

00:00 Every day, developers build new AI tools, and keeping up with all of them takes time. This is our weekly project update video where we cover the top trending AI agent projects this week, so you can find what is worth using in one place. You will discover trending AI tools and projects across creative work, coding, and automation without wasting time.

00:20 Let's get started. >> Before we jump into today's project updates, here's a quick announcement for everyone. We've launched a brand new YouTube channel called AI Agent Studio dedicated entirely to AI Agent projects, tutorials, and tools. So, if you're interested in staying up to date with the latest AI Agent open source projects, learning how to build your own agents, or exploring cuttingedge agent frameworks, make sure to check it out.

00:48 Subscribe now to get weekly videos, in-depth guides, and realtime project breakdowns. The link is right there in the description. Don't miss it. All right, let's get into today's video. >> Project number one, Miniax Design, a local AI crew for creative production. Miniax Design is a desktop creative platform that runs a crew of AI agents from one command center.

01:09 It solves the tool switching and fragmented pipeline that stretch a video or campaign across many separate apps. You describe an idea or drop in a brief and a main agent reads your intent, breaks the job into tasks, and matches the right model. Then four dedicated agents work in parallel on copy, images, video, and voice over. Script, storyboard, video, music, and editing all sit on one canvas whose nodes connect on their own.

01:35 And at key checkpoints, the agent pauses for your direction before continuing. You save workflows as reusable skills or install others, so it learns your style and local files stay on your machine, exportable to ProTools in one click. It runs on Mac and Windows and suits short dramas, e-commerce content, and ads. It is built for creators and studios.

01:57 Download it and direct your first production free. Project number two, Pick Aloud. Turn spoken feedback into an agent plan. Aloud is a Mac app that turns you talking through your app into a clear plan your coding agents can run. It solves the way vague feedback loses context where saying this button leaves an agent guessing. You hit record and talk through the build while it captures your voice, screen, and transcript together with the recorder kept out of the frame.

02:24 It then rewrites the transcript to what you actually meant, drops the ums and the calls you reversed mids sentence, and suggests a screenshot wherever you said this, here, or that button. It groups the result into tasks, asks a quick question when something is ambiguous, and exports one self-contained file you drop into clawed code, cursor, or codeex.

02:44 Transcription runs on device with whisper. It is built for people directing agents to build software. Download it for Mac and record your first pass. Project number three, Checkexum AI, autonomous testing that generates and heals suites. Checksum is a continuous quality platform that generates, runs, and heals your test suite alongside CI/CD. It solves the bottleneck where AI writes code far faster than teams can test it and where brittle tests break and stall releases.

03:13 Unlike a co-pilot, you prompt back and forth. It works as a background agent. You hand it a task like generate 200 tests and it goes off and returns finished production ready playwright code you own. It runs on every commit, pull request, and deployment. And three agents cover the work. An endtoend agent for full user journeys, a CI agent that tests each specific change, and an API agent across thousands of endpoints.

03:38 When your app changes and tests break from shifted selectors or flows, it fixes most on its own, so the suite stays green with little maintenance. It is built for engineering and QA teams shipping fast. Request a demo to see it work. Project number four, Nest Shape. Aentic IDE for designers and programmers. Shape is an Aentic IDE that brings design, code, Git, and AI chat into one desktop app.

04:01 It solves the jumping between editor, design tool, GitHub, terminal, and AI tabs just to ship a single feature. You write in a fast code editor with a built-in AI agent working directly in your repo, and you edit your running app visually rather than hunting through code. Live previews show changes as you make them. Tailwind and modern languages are supported first class, and Git is managed without leaving the editor.

04:25 It is built for designers and programmers who want one workspace. Download it for Windows and build your first feature. Project number five, Bit Drift. Agentic Realtime Observability for mobile apps. BitDrift is an Agentic mobile observability platform that lets AI agents query what users do inside your app in real time and act on it. It solves the blind spot where mobile teams wait 10 days for an app release to change what they log and where crashes hide the frozen frames and silent failures that actually drive churn.

04:54 The lightweight Rust SDK for iOS, Android, and React Native writes rich telemetry to a fixed-size ring buffer on the device, and a control plane decides in real time what to keep and upload, so you log everything locally and send only what matters with no per log overage tax. Workflows trigger a session capture when a sequence stalls, and privacy conscious replays reconstruct the journey alongside logs, network, and app state without exposing personal data.

05:24 Its public API and BD skills let agents triage and investigate on their own. It is built for mobile engineers. Integrate the SDK and start with free crash reporting. Project number six, Zoho Click 7.0. Team messaging with AI and connected agents. Zoho Click is a team communication app whose latest release keeps search, meetings, automation, and AI in one window.

05:47 It solves the constant tab switching and repetitive work that break a team's focus during the day. You pick the AI model behind your workspace, choosing Claude, Gemini, ChatgPT, Cohhere, or Zoho's own Zia, and use it to summarize videos, audio, and recordings. Transcribe voice and run custom prompts written in your organization's tone. A no code drag and drop builder creates bots and guided workflows that reach external systems through web hooks and MCP support plus LLM optimized REST APIs.

06:18 Let agents like Claude and Cursor complete real work inside click with one instruction. Governance keeps AI features off by default under admin control with scoped external access and message reporting. Meetings add live captions, summaries, and availability views. It is built for teams and enterprises. Sign up free and explore the release. Project number seven, Whisper Flow.

06:42 Speak naturally. Get polished text everywhere. Whisper Flow is a voicetoext tool that turns your speech into clean writing across any app. It solves the slowness of typing and the messy output ordinary dictation leaves behind. You talk the way you normally would, and Flow transcribes and autoedits as you go, placing formatted text at your cursor with filler words and typos already gone.

07:06 It picks up your names and jargon in a personal dictionary, turns phrases you repeat into voice snippets, and recognizes over a 100 languages. It runs natively on Mac, Windows, iPhone, and Android, carrying your style and settings across devices, and works inside tools like cursor, VS Code, Slack, and email. It is built for anyone who spends the day at a keyboard.

07:29 Download it free and start talking. Project number eight, Finn, AI agent for support, sales, e-commerce. Finn is an AI customer agent from Intercom that works across support, sales, and e-commerce through the whole customer journey. It solves the round-the-clock flood of questions and tickets that human teams cannot cover alone. It runs on Apex models trained on billions of real customer interactions, so it reasons through multi-step problems, updates accounts, and handles payments and refunds rather than just replying.

07:58 It pulls live data and takes actions through APIs. Data connectors or MCP works natively across voice, chat, email, and social in dozens of languages and plugs into help desks like Intercom, Zenesk, Salesforce, and Freshes. You shape its tone, knowledge, and behavior. Test changes before they go live and pay only for conversations it resolves. It is built for support and CX teams.

08:23 Start a free trial or watch a demo. Project number nine, Origin by Cursor. Code hosting with agents in every repo. Origin is code hosting from cursor that puts your repos, pull requests, and AI agents in one place. It solves the split where your code lives in one tool and your agents in another so context is lost between them. You create a repo in the codebase tab, install the CLI, and push to host it on origin.

08:49 Or connect GitHub and sync existing repos that update in real time while GitHub stays the source of truth. Every repo has pull requests with a timeline, commits, checks, and diffs. And on synced repos, comments and reviews flow both ways within seconds. Because agents sit in every repo, you ask questions about the code you are browsing, and cursor answers, makes changes, updates the pull request or pushes a branch.

09:14 App extensions connect Versal for preview deployments and depot or build kite for CI. It is built for developers working at agent scale. Name your codebase and create your first repo. Project number 10 fihost agents and clueing shared knowledge layer that runs agents. Clueing is a knowledge layer that captures your context and runs hosted AI agents on it for the whole team.

09:35 It solves the scatter of research across tabs, docs and notes and the infrastructure gap that keeps capable agents from doing shared work while you browse. One click saves any insight from a web page, PDF, video or social post with the source and your highlight attached. Every AI model, your claude code, and other agents run on Clueing's infrastructure, so you publish websites, dashboards, and apps in one click on your own domain with secure HTTPS.

10:02 Teams share more than documents, living context, dynamic skills that run on that context. Connectors with encrypted secrets, tasks, and the agents that carry them out, all from one brain with roles and model access controls. It is built for teams and professionals doing research and building. Start free and capture your first insight. Project number 11, Expert Chase 2.0.

10:25 Personal AI that runs your everyday life. Expert Chase is a personal productivity app whose AI called EYE organizes every part of your everyday life in one place. It solves the scatter of running your days across separate to-do, calendar, notes, food, sleep, habit, finance, and fitness apps. EE draws on your real life data and answers from it, not with generic replies.

10:51 So, it stays specific to your world and learns from it to keep you organized and ahead. Anything you can do in the app, EE can do for you. And it connects to Apple and Google through calendar, reminders, tasks, and health. So, your existing schedule and health data flow in. One subscription covers unlimited AI plus those integrations. It is built for anyone who wants everyday life in a single place.

11:14 Download it free on iOS or Android and get started. Project number 12, Monsbyamin Marketplace where AI agents work together. Monsby AI agent marketplace where you discover, run, and build specialized agents for work, creativity, learning, and fun in one place. It solves the scatter of hunting for a different tool for every task. Each built to excel at just one thing.

11:37 You browse purpose-built agents across research, business, design, media, finance, legal, and development. From deep research briefs, and document analysis to music, video, and image generation. In the workspace, you start a group chat with several agents at once, assigning tasks and gathering ideas the way you message co-workers on Slack, so they collaborate on one job.

12:02 In studio, you create your own agent by defining the workflow and connecting your data, then publish it to the marketplace with no machine learning expertise needed. It runs on credits. It is built for anyone who wants specialized AI on demand. Explore the agents and start building your own. Project number 13, victor.com AI employee that lives in Slack.

12:21 Victor is an AI employee that lives in Slack and Microsoft Teams and does the work you cannot get to. It solves the pile of tasks too small to hire for yet too timeconuming to do yourself. You add it from the app directory like a new teammate. Connect your tools and assign a job in a thread. It has its own computer in the cloud where it writes and runs code.

12:42 So it takes a task from research to shipped delivering reports, dashboards, code, and campaigns, then updating your tools. It connects to over 3200 apps like HubSpot, Notion, Linear, Google Drive, and Stripe. learns your processes and preferences over time and stays autonomous but supervised, proposing the next step and asking before anything it cannot undo with approval gates and per person permissions.

13:07 It is built for teams that want to run leaner. Start free and assign your first job. Project number 14, Voss 2.0 local AI voice studio without limits. Voss is a local AI voice generator studio that runs texttospech and audio processing on your own machine. It solves the per character fees, credit meters, and privacy trade-offs of cloud voice tools. You write, cast, generate, and master in one place.

13:34 Choosing from over a 100 natural voices across categories. Cloning a voice from about 15 seconds of audio or designing one from a text description on Pro. It voices multis-speaker scripts, rrenners just the line you edit rather than re-recording, and masters to platform loudness like Spotify and Apple in one click. Because it is AI native, you connect Claude, Chatg GPT, Gemini or any agent through its skills and let it run the whole studio and nothing leaves your computer.

14:03 Plans are flat and unlimited with pro adding over 600 languages. It is built for podcasters, audiobook authors, and video and game creators. Download it and start a free trial. Project number 15, Research AI, reproducible science workspace run by agents. Research AI is an AI workspace where agents run your whole scientific workflow in one conversational place.

14:25 It solves the constant switching between chat bots, Python, R, reference managers, search tools, and a word processor that scatters research and breaks reproducibility. You describe what you need and agents handle literature search, data acquisition, statistical analysis in Python and R, visualizations, editing, and scientific writing. each step traceable and reviewable.

14:49 Every workflow runs in an isolated cloud sandbox separate from their servers, so your files stay private and each result is fully reproducible with no software to install. It connects to 60 scientific sources like PubMed, Semantic Scholar, Aex, Unirot, and Clinical Trials. Gov and its rich editor exports formatted word documents with tables, figures, and academic pros.

15:13 It focuses on biosciences and adapts to your domain in several languages. It is built for scientists and research teams. Start free and turn your data into publishable results. Project number 16. Tines governed workspace for AI agents and automations. Tines is a secure workspace where every team builds AI agents, apps, and automations. While IT and security keep oversight, it solves the shadow IT risk that spreads when people build alone.

15:42 committing keys to repos and running apps on personal accounts. Teams self-s serve on a govern path so security sees what exists, who made it, and what it can touch. Its agents run sensitive workflows like alert triage, IT deployments, finance approvals, and hiring, surfacing suggested fixes for anomalies while you decide what actually goes live. It stays vendor agnostic across your stack, and lets you choose where and how your models run, so data stays under your control.

16:08 It is built for security, IT, and the teams that rely on them. Sign up free or book a demo. Project number 17, Antibbot, AI community manager for your Discord. Antibbot is an AI community manager for Discord that starts conversations, remembers members, and keeps a server lively. It solves the two things ordinary bots miss. Quiet channels and clunky slash commands nobody wants to memorize.

16:33 You set it up in three steps. Choosing a name, avatar, and personality. or writing your own, then members just talk to it in plain English rather than digging through docs. Its agentic tools cover moderation, polls, music, web research with cited sources, and catch me up recaps that link back to what happened. It remembers each member's interests and inside jokes, so check-ins feel personal.

16:56 And when a channel goes silent, it starts a conversation on its own based on what members like. Auto threading, daily recaps, rule flagging, and an admin dashboard keep things organized, and its AI runs on a shared token pool. The community can top up. It is built for Discord community owners. Add it to your server free. Project number 18, Cherry Blossom, AI agent that designs your PCB.

17:20 Cherry Blossom is an AI agent platform for electronic design that takes a circuit board from plain language requirements to fabe files. It solves the way ordinary AI PCB tools fall apart on hard boards and hand the complexity back to the engineer. You describe the board you want and the agent handles the physics, component selection and layout end to end, producing an editable schematic, a routed PCB and manufacturing files.

17:44 It runs design rule and electrical rule checks inside its own cloud workspace as part of each turn. So every command is in the transcript and you watch a failed check fail rather than trusting a report. runs finish in a sandbox even if you close the laptop and boards export as an open format plus a kyad project with gerbers and bomb so nothing is locked in.

18:05 It works on top of your existing EDA or standalone with free kyad priced on usage with no per seat cost. It is built for electrical engineers, teams and hobbyists. Describe a board and start free. Project number 19 kisdr video AI rep that qualifies website visitors. Clara is a face-to-face video AI sales development rep that turns website traffic into pipeline.

18:30 It solves the leads lost when interested visitors hit a form or land outside business hours and cool off before a rep replies. Clara engages visitors in real time on your site or inapp. Running a personalized video demo that adapts to each person's industry role and needs, handling objections and keeping the conversation on value. It reads intent, asks the right questions to qualify high-v valueue prospects, and books meetings on its own, working around the clock in major languages.

18:59 You create the rep by uploading or generating a face, train it on your decks, scripts, and FAQs so it speaks your brand and launch on your site within a day. It plugs into your CRM to sync contacts and log calls with enterprisegrade security. It is built for sales and marketing teams. Launch a demo and put CL to work. Project number 20, Taku AI Marketplace that assembles AI into workflows.

19:23 Taku is a marketplace and builder that turns AI apps, agents, and skills into one-click tools you can run and remix. It solves the gap where you find a powerful AI setup, but never actually get to use it, and where your own best workflow dies in a doc nobody reads. You browse apps in the marketplace and save them or describe a goal in plain words and Taku assembles the right apps, agents, and skills into a stack made for you.

19:51 Live with no config. When you build something good, you package it as a stacks that others install, run, and make their own with nothing to install on their end. Creators publish once and earn every time their stacks runs. So, a workflow travels through the community. It is built for anyone who wants working AI without wiring it together. Download it and borrow your first tool. Thanks for watching.