Connect separate applications into automated workflows using natural-language instructions and AI agents that understand tasks, make decisions, and select tools.
Behind this: 8 build steps · 1 tool and how each is used · how to validate demand · 2 things the video never answers.
Provide a hardware interface that lets developers monitor and control several coding agents without repeatedly switching windows.
Behind this: 6 build steps · 1 tool and how each is used · how to validate demand · 2 things the video never answers.
Operate an AI account executive that responds to inbound buyers, conducts live demos, qualifies prospects, follows up, and closes smaller deals.
Behind this: 9 build steps · 1 tool and how each is used · how to validate demand · 3 things the video never answers.
Provide agents with domain-based inboxes and MCP access so they can create accounts, send email, read inboxes, reply, and operate with approval gates.
Behind this: 11 build steps · 1 tool and how each is used · how to validate demand · 2 things the video never answers.
Let users describe website changes in natural language while agents design, manage content, and create custom interactions directly on an editable canvas.
Behind this: 9 build steps · 4 tools and how each is used · how to validate demand · 2 things the video never answers.
Capture company decisions, owners, plans, and drift signals from work tools, then expose current shared memory to AI agents through MCP.
Behind this: 8 build steps · 1 tool and how each is used · how to validate demand · 2 things the video never answers.
Observe workflows in ERPs, mainframes, desktop applications, and internal portals, then convert them into stable, governed MCP tools.
Behind this: 11 build steps · 1 tool and how each is used · how to validate demand · 2 things the video never answers.
Create autonomous digital employees from plain-language role descriptions that plan and execute work around the clock.
Behind this: 10 build steps · 1 tool and how each is used · how to validate demand · 2 things the video never answers.
Turn spoken thoughts and meetings into linked visual cards that preserve reasoning, decisions, risks, and unanswered questions.
Behind this: 10 build steps · 1 tool and how each is used · how to validate demand · 2 things the video never answers.
Monitor airline fares continuously and send travelers fast, personalized alerts about unusually low fares or possible pricing errors.
Behind this: 9 build steps · 1 tool and how each is used · how to validate demand · 2 things the video never answers.
Capture and connect material encountered while browsing so users and AI agents can search and reuse grounded research context.
Behind this: 8 build steps · 1 tool and how each is used · how to validate demand · 2 things the video never answers.
Audit an existing software product, create a dependency-mapped backlog, and use agents to implement approved work with human pull-request review.
Behind this: 10 build steps · 6 tools and how each is used · how to validate demand · 2 things the video never answers.
Run virtual wind-tunnel simulations on uploaded 3D models without installing traditional aerodynamics software.
Behind this: 10 build steps · 2 tools and how each is used · how to validate demand · 2 things the video never answers.
Build a live infrastructure knowledge graph that groups alerts, traces blast radius, suppresses known noise, and diagnoses root causes.
Behind this: 11 build steps · 8 tools and how each is used · how to validate demand · 2 things the video never answers.
Give agents persistent memory and document retrieval through native MCP tools, with hybrid search and cited passages.
Behind this: 11 build steps · 3 tools and how each is used · how to validate demand · 2 things the video never answers.
Provide on-demand email inboxes that applications and AI agents can use through REST APIs, SDKs, and MCP.
Behind this: 7 build steps · 1 tool and how each is used · how to validate demand · 2 things the video never answers.
Generate editable, rigged, and animatable 3D assets from text prompts or reference images in a browser-based workflow.
Behind this: 10 build steps · 1 tool and how each is used · how to validate demand · 2 things the video never answers.
Ingest signals from a startup's tools, understand operational events, and dispatch specialized agents that produce real business artifacts.
Behind this: 10 build steps · 7 tools and how each is used · how to validate demand · 2 things the video never answers.
Turn a hiring brief into a candidate-scoring rubric, search live people and company data, refine results through recruiter feedback, and enrich and route shortlisted candidates.
Behind this: 12 build steps · 4 tools and how each is used · how to validate demand · 1 more real example · 2 things the video never answers.
Manage agent teams like employees by assigning responsibilities, metrics, and deadlines while approving work and gradually granting autonomy based on performance.
Behind this: 13 build steps · 7 tools and how each is used · how to validate demand · 2 things the video never answers.
Searchable transcript of Top AI Agent Projects : Albato AI, River, Verse, Breadcromb & Agently — ManuAGI - AutoGPT Tutorials (17:30). 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 Everyday developers build new AI tools and keeping up with them takes time. This is our weekly project update video where we cover the top trending AI agent projects this week so you can see what is worth your attention. You will discover trending AI tools and projects in one place without wasting time. Let's get started. >> Before we jump into today's project updates, here's a quick announcement for everyone.
00:22 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. Subscribe now to get weekly videos, in-depth guides, and real time project breakdowns.
00:48 The link is right there in the description. Don't miss it. All right, let's get into today's video. >> Project number one, Albato AI. AIdriven no code platform connecting your apps. Albato is an AIdriven no code integration platform that makes your separate tools work together as one system. It solves the manual repetitive work of moving data between apps by hand.
01:11 You describe what you want automated and Albato Copilot asks the right questions, maps the fields, builds the logic and launches the workflow for you in any language. Its AI agents go further, understanding a task, making decisions, and picking the right tools to finish it. It connects over 1,000 apps like Gmail, HubSpot, Slack, and Telegram. Works with any API, and brings models such as Chat, GPT, Gemini, and Claude straight into a flow.
01:36 Built-in tools handle routing, filtering, and data control. It is built for teams and individuals who want automation without code. Start free and build your first workflow. Project number two, Codeex Micro. A hardware command center for coding agents. Codeex Micro is a compact mechanical keypad built by OpenAI's supply code with work louder as a physical control surface for agentic coding work.
02:00 It solves the constant window switching that comes with running several AI agents at once. Its agent keys light up with live RGB status from codecs. So, a glance tells you which agent is thinking, running, waiting, or done. A joystick flick triggers common workflows like reviewing a pull request, debugging an error, or refactoring code, while command keys put accept, reject, pushto talk, and new chat under your fingers.
02:23 A rotary dial adjusts the reasoning level in the moment. It connects over Bluetooth or USBC on Mac and Windows and pairs with chat GPT codecs and work louder input. It is built for developers shipping with agents. Check the shop for availability. Project number three, River, an AI account executive that closes deals. River is an AI sales agent that acts as a digital account executive for inbound deals.
02:47 It solves the pipeline lost when interested buyers are handed a calendar link and cool off before anyone calls them. Its agent joins a live video call within seconds, shares its screen, and walks the buyer through your product in several languages. For smaller deals, it goes end to end, sending the proposal, following up over email and WhatsApp, and getting the contract signed.
03:06 For larger ones, it qualifies and demos, then hands your team a warm prospect with transcripts and CRM ready notes. It runs around the clock across time zones, and you pay only on closed deals. It is built for sales and marketing leaders growing without extra headcount. Start a live demo and see it work. Project number four, Nitro Send. Full stack email built for AI agents.
03:28 Nitro Send is a full stack email platform built for AI agents to operate on their own. It solves the problem that ordinary email services are hostile to agents needing dashboards, human signups, and giving an agent no address of its own. You point your agent at a single skill file, and it reads the whole platform, creates the account, connects your domain, sorts billing, and sends its first email without a human step.
03:51 It connects to Claude, Codeex, Cursor, Chat, GPT, and Gemini over MCP. So agents send marketing and transactional mail, then read, search and reply from real inboxes on your own domain. Routine replies are handled and anything sensitive is forwarded with approval gates where you choose and your team keeps Gmail as usual. It is built for people running agents at scale.
04:11 Point your agent at the skill and start free. Project number five, Framer Fisha AI website builder with agents on canvas. Framer is an AI website builder where design agents work directly on your canvas. It solves the scattered workflow of designing, managing content, and writing code in separate tools. You describe a change and the design agent generates and refineses it in place with every edit visible, editable, and reversible.
04:35 A CMS agent sets up and organizes your collections so content and design stay in sync, while a code agent turns ideas into custom effects and interactions. External agents like Claude Code, Codeex, or Cursor can drive your site from Slack, the terminal, or a pull request. Hosting, performance, SEO, localization, analytics, and AB testing are built in.
04:58 It is built for designers, teams, and creators. Get started free and publish your site. Project number six, in parallel MCP, shared company context for every AI agent. In parallel MCP is a context layer that gives every AI tool in your stack the same live picture of your business. It solves the problem of assistants that write fluent answers without knowing what your team decided last week or what is slipping now.
05:22 It captures decisions, owners, plans, and drift signals from your calendar, email, and meetings. Then exposes that shared memory over MCP so agents answer from reality instead of a stale doc. You paste one workspace URL into clawed chat GPT, Copilot, cursor, or coding and automation tools, and each workspace stays a separate perimeter with its own permissions, so nothing leaks across contexts.
05:44 Data is EU hosted and never used to train models. It is built for teams across product, sales, finance, and leadership. Connect your workspace and start free. Project number seven, Graph AI makes legacy software usable by agents. Graph AI is enterprise infrastructure that turns workflows inside old software into govern tools for AI agents. It solves the gap where ERPs, mainframes, desktop apps, and internal portals have no API, so agents simply cannot reach them.
06:13 It observes a real workflow and maps the screens, inputs, transitions, and side effects, then compiles a typed contract with policy boundaries, and conformance tests. An independent check verifies the actual state in the source system, so the adapter cannot vouch for itself. Your agents then call one stable MCP tool instead of rediscovering the interface each time.
06:34 With approvals, lease privilege access, and a full audit log, it is built for enterprises running systems they cannot replace. Bring one blocked workflow and start the assessment. Project number eight, Verse. Build autonomous AI employees from one prompt. Verse is a no code platform for building autonomous AI employees that run work around the clock.
06:53 It solves the ceiling on output that small teams hit when every recurring task still needs a person. You describe a role in plain language, grant access, and an employee deploys in minutes, planning and acting on its own within that role. Each one gets its own email, phone, virtual card, computer, and wallet so it can browse, write, and run code. Use over a thousand connectors or reach any MCP server, API, or custom tool.
07:17 In shared spaces, employees collaborate and delegate to each other, and memory keeps context between runs. It is built for founders, agencies, and lean teams scaling without headcount. Describe a role and hire your first employee. Project number nine, weave. Turn spoken thinking into a live map. Weave is a voice-driven thinking tool that turns what you say into a live visual map.
07:39 It solves the way ideas and meetings scatter into messy notes that lose the reasoning behind them. You hold a key or leave the mic open and talk, and your sentences become linked cards within seconds with half-formed thoughts shown as dashed guesses that firm up as you decide. When you pause, weave tidies the board and raises the questions you have not asked yet.
07:59 And any card can be opened into its options and risks. It captures whole meetings too. Drawing who said what, the decisions and the open questions as people talk. Boards are saved, replayable, sharable by link and exportable to claude, markdown, PNG or SVG. It is built for anyone who thinks better out loud. Create your first board and start talking.
08:19 Project number 10, Flight Glitch. AI alerts for airline pricing mistakes. Flight Glitch is an AI powered flight deal service that monitors airline pricing errors around the clock. It solves two problems with cheap fair alerts. The best ones vanish within hours, and most are irrelevant to where you actually fly from. You pick your home airports, and it sends fast alerts when unusually low fairs or possible pricing mistakes appear on your routes.
08:43 Each deal comes with an AI breakdown of the price, the route, the pros and cons, and how far the fair deviates from normal, including the risk that an airline voids the booking. An AI concierge answers questions about a trip, and direct links take you to booking options. It is built for travelers who want to judge a deal fast before it disappears. Pick your airports and start getting alerts.
09:03 Project number 11. Breadcrom AI browser that builds your knowledge graph. Breadcrumb makes trace an AI native browser that turns everything you read into a private knowledge graph. It solves the way research scatters across tabs, documents, and chats, leaving your AI tools to start from scratch every session. As you browse and work, Trace captures and organizes that material into a connected graph you can search.
09:28 So the browser remembers what you have read and understands what you are working on. That stored context is then available to AI agents, giving them real grounding to help you research, write, and automate tasks. It is local first by default and free to download. So privacy is not the price of memory. And the graph can be shared with your team. It is built for researchers, writers, and knowledge workers.
09:48 Download trace and let your browser remember. Project number 12. Son of aentic engineering service that builds on shipping. Son of is an agentic engineering service that ships production code into your existing product. It solves the backlog that keeps growing because your team or the developer who left cannot get through it. You connect your repo along with Jira, Linear, ClickUp, Notion, and Slack.
10:10 And an agentic audit reads all of it in a day, reporting security gaps, tech debt, and architecture issues in plain English. From that, it writes a dependencymapped backlog with story points and dollar costs. A senior tech lead reviews each ticket. You approve what matters. Agents run the work around the clock, and a human signs every pull request before it merges.
10:28 You are built only for code that reaches production. It is built for founders and teams shipping more than they can staff. Connect your repo for a free audit. Project number 13, Ventora, virtual wind tunnel running in your browser. Ventora is a browserbased computational fluid dynamics tool that runs a virtual wind tunnel on any 3D model. It solves the cost and setup burden of traditional aerodynamics software that needs installs and license servers.
10:54 You upload an STL OBJ step or GB file, set wind speed, direction, fluid and turbulence model, then watch streamlines pressure fields in wake form as the solver iterates. It reports lift, drag, and moment coefficients, Reynolds and mock numbers, and surface maps. Exportable to CSV or PDF with charts and side-by-side run comparison. An AI arrow assistant explains why drag is high or how to add lift running on clawed and a parallel open foam backend is benchmarked against published wind tunnel data.
11:24 It is built for aerospace, automotive, and drone engineers. Upload a model and run your first simulation free. Project number 14, Alert Grouping by Dr. Droid. Self-learning AI agent for incident response. Dr. Droid is a self-learning AI agent for site reliability that turns your stack into a living knowledge graph. It solves the noise and guesswork of on call work where alerts pile up and engineers hop between dashboards to find the cause.
11:51 You grant readonly OHOF access to cloud code, CI/CD and observability tools and it crawls metrics, logs, traces, repos, runbooks and wikis to map which repo belongs to which service. dashboard and pods. When an alert fires, the graph traces the blast radius in seconds, groups related alerts, suppresses known noise, and diagnoses root cause. It remembers every investigation, so repeat incidents resolve with fewer steps.
12:18 It connects with AWS, Data Dog, Graphfana, Pedager, Duty, Sentry, Slack, and Jira, and can run inside your own network. It is built for engineering and on call teams. Connect your stack and generate your graph. Project number 15, Kit for AI. Memory and knowledge layer for AI agents. Kit for AI is a memory and knowledge layer that gives AI agents lasting context.
12:41 It solves the problem of assistants that start every session blank and cannot read the documents your work lives in. Your agent calls remember, recall, and search as native MCP tools. So, preferences and decisions carry across conversations. You drop in a PDF, Word file, spreadsheet, scanned image, web page, or YouTube link, and it comes back as clean markdown, chunked, embedded, and searchable with structured JSON if you need a schema.
13:04 Hybrid semantic and keyword retrieval returns cited passages instead of whole documents, cutting tokens sharply. It works with claw, cursor, and any MCP client over one API, and your data stays encrypted and private. It is built for developers building agents and rag pipelines. Start free and give your agent memory. Project number 16, InBix. Open-source Cloudflare native email API platform.
13:28 Inbix is an open-source Cloudflare native email API platform for programmatic inboxes. It solves the work of standing up mail infrastructure yourself when an app or agent needs a real address to receive it. You generate inboxes on demand and receive emails through it, then build automation on top. It exposes REST APIs and SDKs for code, plus MCP, so AI agents can use the same inboxes directly as tools.
13:51 Being open- source, the whole stack stays inspectable and self-hostable on Cloudflare. It is built for developers wiring email into products and agents. Explore the docs and spin up your first inbox. Project number 17, V2 Fun. AI 3D creation from text or image. V2 Fun is a browser-based AI 3D creation platform that turns text prompts or images into finished 3D assets.
14:16 It solves the fragmented pipeline where modeling, texturing, rigging, and animation each need separate software and years of skill. You describe an idea or upload reference images, optionally including front, side, and back views, and its AI generates a structurally accurate model. A prompt optimizer enriches your input with perspective. PBR materials and texture detail, while smart rettopology returns clean editable quad meshes.
14:43 Automatic rigging turns a static model into an animatable character, and its motion capture reads human movement from ordinary video with no suit or hardware. Assets export to FBX and GB for game engines or 3D printing. It is built for indie developers, designers, and creators. Start creating your first 3D model. Project number 18, Agently. A company brain that runs your stack.
15:06 Agently is a multi- aent platform that builds a company brain across your tools and then acts on it. It solves the gap left by assistants that answer questions but never do the work. You connect your stack with OOTH and the brain ingests signals from every tool so it knows a charge failed, a deal slipped, and a ticket is waiting. An orchestrator called Jarvis spins up specialized agents for research, revenue, growth, support, and ops, roots the work, and ships.
15:31 It outputs land as real artifacts, including Docs, DEX, Sheets, and Pages. It connects to over a 100 tools through MCP, including Slack, Linear, Notion, Stripe, HubSpot, and GitHub with guard rails, and audit receipts. It is built for small startup teams. Connect your tools and put the brain to work. Project number 19, Crust Data Recruiter. Build your sourcing twin inside Claude.
15:53 Crust Data Recruiter is a set of recruiting skills and an MCP server that turn Claude into a sourcing partner. It solves the flood of near miss candidates that keyword-driven tools return when a brief calls for judgment. You paste the hiring manager call and job description, and it builds a rubric with your hard gates and scoring, then searches live people and company data covering over a billion candidate profiles.
16:15 It returns batches of 20 to 25 scored candidates, each with a oneline reason, and you correct it in plain language until the next batch fits. It then enriches only your short list with verified email and phone, writes to an ATS like LockXO or GEM, and triggers outreach in your voice. It is built for recruiters and search firms. Add the connector and claude and brief it.
16:37 Project number 20, Yagny. Agent teams you manage like people. Yagni is a platform for proactive agent teams that you manage the way you manage employees. It solves the pile of unreed drafts that ordinary agents create. Since trust never arrives with a bot, you must babysit. You give a team its responsibilities in plain language, one number it is measured on, and commitments with real end dates.
16:58 It proposes work, you approve or edit, and your edits become a playbook that shapes every future draft. Routine reversible work runs itself and returns a receipt from the source. While consequential calls wait for your decision, every team starts in training and earns autonomy rule by rule on its track record. It reads your existing stack, including Slack, Gmail, HubSpot, Stripe, GitHub, and Linear, and ships approved work back there. Start free and build your first team. Thanks for watching. See you in the next video.