10 AionUI Use Cases with Claude Code, Codex & Hermes Agent

Added: 2026-05-10

[00:00:00]You have Claude Code installed. You have Codeex. Maybe Hermes agent, Open Claw, Open Code, a whole roster of some of the most capable AI coding agents available right now. And you're still switching between terminals, losing context, duplicating config files, and manually holding everything together like some kind of AI traffic cop. None of your agents know each other exist. None of them share tools. None of them coordinate. They're all incredibly powerful on their own and completely disconnected from one another. And that gap between what they could do individually and what they could do together is exactly where this video lives. Ion UI is a free open source co-work platform that puts every CLI agent you already have under one unified interface. It is not another chat wrapper. It is not a web UI for a single model. It is a desktop environment where your agents actually work together sharing tooling, coordinating on tasks, running in parallel, and operating continuously even when you're not at your computer. And it does all of this without touching your existing agent setup, without overriding your configs, and without costing you anything. In this video, we are breaking down 10 specific agentic use cases inside IONUI, covering multi-agent orchestration, unified MCP configuration, scheduled automation, remote access, VPS deployment, office document generation, image generation workflows, and more.

[00:01:17]Each use case is built around the agents you already know. Clawed Code, Codeex, Hermes agent, Open [music] Claw, Open Code, and what becomes possible when ionui gives them shared environment to operate in. This is not a setup guide.

[00:01:29]This is a breakdown of what the platform actually unlocks. If you find this video useful, go ahead and hit the like button right now. It genuinely helps this channel reach more people who are deep in the AI agent space and would benefit from content like this. If you're not subscribed yet, now is a good time. This channel covers AI developer tools, open source agents, and the infrastructure behind them on a regular basis. And if you know someone who is already running Claude code or codecs and has no idea ionui exists, share this with them. It will save them a lot of friction. Now, let's get into it. Ion UI stands for ION user interface and it was built by Office AI as an open- source desktop application that sits on top of all the CLI based AI agents you are already using. The core idea is simple. Instead of every agent living in its own isolated terminal with its own configuration and its own disconnected history, ION UI gives them a single shared environment, one interface, one place to manage everything with full visibility into what every agent is doing at any given moment. It is free Apache 2.0 licensed crossplatform on Mac OS, Windows, and Linux. And it stores all of your conversation data locally in a SQLite database. Nothing is routed through their servers. Now the important thing to understand before we get into the use cases is that ion UI is not just a launcher. It ships with its own built-in agent called ion CLI which is a rustbased agent engine that is bundled directly into the app. That means the moment you install ion UI you already have a fully capable agent with file read and write access web search image generation and MCP tool support. No additional CLI tools required. You can start using it immediately with a Google signin for Gemini or by dropping in an API key from any provider you prefer.

[00:03:09]But where it gets genuinely interesting is the multi- aent layer. ION UI autodetects any CLI agents you already have installed. Claude Code, Codeex, Hermes agent, Open Claw, Open Code, Quincode, Goose, Gemini CLI, Augment Code, Cursor Agent, and more than 15 others. It pulls them into the same interface, gives them access to the same MCP configuration and lets them run in parallel sessions with independent context. That is the foundation.

[00:03:34]Everything else in this video is built on a single platform where your entire agent stack actually coexists. Now, let's get into the use cases. The first use case is the one that separates ionui from every other agent interface out there, and it is called multi-agent team mode. This is where you stop thinking about your agents as individual tools and start running them as an organized team with a leader agent that receives your instructions, breaks the work down into subtasks, and delegates to teammate agents that execute in parallel underneath it. Here is how it actually works. You give ion UI a highle instruction, something like build out the backend for this feature or research and draft a competitive analysis for this product. The leader agent takes that instruction, figures out what needs to happen and decomposes it into discrete subtasks. Those subtasks get assigned to teammate agents that start executing simultaneously. They are not waiting for each other. They are not sharing a context window. Each one is running independently with its own session working on its assigned piece of the project. All the agents in a team share the same project folder. What ION UI calls the team isolated workspace. As teammates complete their work, they write their outputs to a shared taskboard and results flow back to the leader through an async mailbox. The leader aggregates everything and surfaces the final output to you. The whole time this is happening, every agent has its own permission dialogue with a sidebar badge for pending approvals. So you can see exactly what each one is doing and approve or reject actions without the process blocking.

[00:04:58]The confirmed leader backends in team mode are Claude Code, Codeex, Hermes agent, Gemini, Snow CLI, and ion CLI. So in practice, you could set Claude code as the brain, the leader that handles decomposition and reasoning, and have codecs and open code running as teammates underneath it, executing the actual generation tasks in parallel.

[00:05:17]That combination alone changes the scale of what you can realistically hand off to your agent stack in a single session.

[00:05:23]A project that would have required you to manually coordinate three separate terminal sessions now runs as one orchestrated workflow from a single instruction. If you're running more than one AI agent right now, you already know the MCP configuration problem. You set up a GitHub MCP server for Cloud Code, then you have to set it up again for Codeex, then again for Hermes agent, then again for Open Claw. Every agent has its own config file. Every time you add a new MCP server, you're editing four files instead of one. Every time something breaks, you're debugging four separate configurations. It is one of those friction points that does not seem like a big deal until you are actually living with it every day and then it becomes genuinely exhausting. Ion UI solves this at the infrastructure level.

[00:06:06]You configure your MCP servers once inside ionui and every agent that is connected to the platform picks it up automatically. Add a file system MCP, a browser MCP, a GitHub MCP, a posters MCP, it does not matter. You set it up in one place and it applies everywhere.

[00:06:23]change it in one place and the change propagates everywhere. There is no per agent configuration to maintain, no duplicate files to keep in sync and no debugging across multiple separate setups. The deeper shift here is architectural. When your MCP layer is centralized, it stops being an agent specific configuration detail and starts being a shared capability layer that your entire agent stack can draw from.

[00:06:45]Every agent in ION UI, whether that is Claude Code, Codeex, Hermes agent or the built-in ion CLI is operating with the same tool surface. That means when you add a new MCP server to extend what your agents can do, all of them benefit immediately, not just the one you happen to configure first. The impact of this use case scales directly with how many agents you're running. If you're only on cloud code, centralized MCP configuration is a convenience. If you're running clawed code alongside codeex, Hermes agent and openclaw simultaneously, which is exactly where this platform is pushing you, it becomes essential infrastructure. And as MCP continues to establish itself as the standard tool protocol across the AI agent ecosystem, having a centralized management layer for it is only going to matter more over time. Most people who are running multiple AI agents today are doing it sequentially. You finish a task with clawed code, close it out, open Hermes agent, do your research, close that out, open claw, run your automation one after another in a line, waiting for each one to finish before you move to the next. That is not multi-agent work.

[00:07:49]That is just switching tools manually and calling a workflow. Ionui's parallel sessions break that pattern entirely.

[00:07:56]Inside ION UI, you can run multiple agents simultaneously in completely separate sessions. Each with its own fully independent context window, not tabs that share state, not windows that reference the same conversation.

[00:08:08]Genuinely isolated, genuinely parallel execution. Claude Code in one session working through a codebase refactor.

[00:08:15]Hermes agent in another session doing deep research on a competitor. Open Claw in a third session handling file organization and automation tasks. All of them running at the same time, none of them aware of or interfering with each other, and all of it visible from a single interface. The conversation history for every session is saved locally, so you can close a session and return to it exactly where you left off without losing context. If cloud code was midway through a complex task when you needed to step away, that session is waiting for you. Not a blank slate, not a summarized approximation, exactly where it was. The real value here is not just speed. Though the time savings are significant, [music] it is about matching the right agent to each type of work and letting them run concurrently without you having to manually manage the sequencing. A coding agent, a research agent, and an automation agent all have different strengths. Running them in parallel on different aspects of the same project means you're using those strengths simultaneously rather than trading them off against each other. The gap between what you can accomplish in an hour of parallel agent work versus an hour of sequential tool switching is not marginal. It is a fundamentally different output ceiling.

[00:09:21]Every AI agent workflow you have right now is reactive. You open a terminal, you write a prompt, you wait for a response, you do something with the output. The agent exists to respond to you and when you're not there, nothing happens. That is a fine starting point, but it is an artificial ceiling because the whole premise of aic AI is that these tools can operate with real autonomy. Cron scheduling in ionui is where that autonomy becomes practical and continuous rather than theoretical.

[00:09:48]Ion UI has a native scheduling system built directly into the platform and you configure it entirely in natural language. You do not write a chron expression. You do not touch a config file. You describe what you want and when you want it something like every morning at 9, summarize yesterday's GitHub commits and send the result to telegram or every Monday pull last week's analytics and generate a performance report. ion UI converts that description into a chronic expression internally and handles the rest. The task runs on schedule automatically whether you're at a desk or not. What makes this more than a basiculer is that every scheduled task is bound to a specific conversation inside ioni. That means the agent executing the task is not starting cold every time. It has the context of the conversation it was set up in which makes recurring tasks significantly more coherent and useful over time. When the task completes, the output comes back into that conversation window and can also be pushed automatically to Telegram, Lark or Dingtalk, so you receive the result wherever you are. Any agent connected to ion UI can be the executive for a scheduled task. Hermes agent for researchheavy recurring workflows. Open Claw for file management and automation jobs that need to run on a schedule.

[00:10:59]Claude code for morning code review or a nightly repository summary. You assign the agent based on what the task actually requires, not based on what happens to be open in your terminal. The result is an agent stack that is no longer waiting for you to show up and give it something to do. It is working on a schedule, surfacing results, and operating continuously as infrastructure rather than as a tool you have to manually invoke every single time.

[00:11:21]There's a version of AI agent work that is completely tied to your desk. You're physically present, your laptop is open, your terminals are running, and everything stops the moment you walk away. Ion UI has two remote access modes that break that dependency entirely and together they turn your local agent stack into something you can reach from anywhere in the world. The first is web UI mode. Ion UI can run as a standalone HTTP server which means you access your entire agent environment through a browser from your phone from a tablet from another computer from anywhere with an internet connection. The interface is the same when you get on the desktop.

[00:11:55]You get the same sessions, the same agents, the same conversation history.

[00:12:00]Login works via QR code or account and password and it supports LAN access, cross network access, and full server deployment. If you're already hosting on a VPS, which we will cover in the next use case, web UI mode is what gives you persistent browser access to your agents around the clock. The second is messaging app integration. ION UI connects natively with Telegram, WeChat, Lark, and Ding Talk. You pair your ION UI instance with a bot, and from that point on, you can send commands directly from your messaging app. The agent on your machine receives the command, executes it, and pushes the result back to the chat. It is not a simplified version of the agent. It is the same agent, the same capabilities, accessed through a message instead of terminal.

[00:12:40]In practice, these two modes cover different scenarios. Web UI gives you the full interface when you need to do real work remotely, reviewing sessions, managing agents, handling approvals.

[00:12:50]Telegram integration handles the lighter side. Firing off a quick command, checking on a running task, getting a result pushed to you without having to open anything. Together, they mean the gap between you and your agents is no longer measured in physical distance from your computer. The agents are running, they are reachable, and you are in control regardless of where you are.

[00:13:09]Everything we have covered so far, the scheduled automation, the remote access, the parallel sessions, all of it becomes significantly more powerful when your ionui instance is not running on a laptop that closes at the end of the day. VPS deployment is what takes ionui from a desktop productivity tool to a persistent agent infrastructure that operates continuously without any dependency on your local machine. Ionui supports a headless web UI mode, which means it can run as a standalone HTTP server with no display required. You deploy it on a VPS, point your browser at the server address, and your entire agent environment is there, fully functional, always on, accessible from any device. The agents are not paused.

[00:13:48]The cron jobs are not waiting for your laptop to wake up. The telegram integration is live. Everything runs continuously on the server while you go about your day, sleep, travel, or work on something else entirely. From an infrastructure standpoint, Ion UI itself is lightweight. It is built on electron with a SQLite backend. So the memory and compute footprint of the platform itself is not significant. The real resource cost comes from the agents you are running and the API calls they are making which means you can run a very capable ioni deployment on a modest VPS without needing enterprisegrade hardware. A standard 2 to four core instance with 4 to 8 GB of RAM is enough to run ionui alongside openclaw Hermes agent and the built-in ion CLI handling scheduled and on demand tasks simultaneously. The agents that make the most sense in a server deployed context are the ones designed for headless low intervention operation. Openclaw with its tool gateway is a natural fit. It is built for autonomous execution. Hermes agent handles research and automation workflows that do not need constant human input. Ion CLI handles scheduled tasks through the cron system. And when you do need to intervene or issue a new instruction, web UI and telegram are right there. This is the architecture that turns your agent stack into infrastructure. Not a set of tools you open when you need them, but a system that is running, working, and delivering results whether you're there or not. If you need a reliable VPS to host your ion UI setup, Hostinger is what I recommend.

[00:15:12]The link is in the description below.

[00:15:14]Most developerfacing agent tools have a very narrow definition of what a useful output looks like. Text, code, maybe a markdown file, and that works well for a certain category of tasks. But there's a large portion of real work that ends up in a PowerPoint deck, a Word document, or a spreadsheet. And most agent platforms just drop you off at the edge of that world and expect you to handle the formatting yourself. Ion UI closes that gap with a set of built-in office assistants that generate actual editable office files natively inside the platform. These assistants are powered by Office CLI, which is a document generation layer built directly into ionui. The outputs are real files, pptx, docs, and xlsx, not PDF exports, not screenshots of rendered content, not markdown that you have to manually pay somewhere else. Actual editable office documents that open in PowerPoint, Word, and Excel the same way any other file would. The PowerPoint output even supports morph transition animations, which is genuinely uncommon in any agent tool currently available. The built-in assistant lineup covers a wide range of document types. You have the PPT creator, Morph PPT for threedimensional animated presentations, a pitch deck creator, a word creator, an Excel creator, a financial model creator, a dashboard creator, and an academic paperwriter among others. These are not toy templates. They are full assistants you can direct with natural language and they produce structured formatted documents as output. The interesting angle from an agentic workflow perspective is how these assistants connect to everything else running inside ioni. You're not using the document generation tools in isolation.

[00:16:47]Claude code can generate the data or the structured content upstream. Hermes agent can handle the research that feeds into a report. Open code can produce the code snippets that go into a technical document. And then the office assistant takes all of that and produces the formatted output all inside a single ioni session without you ever leaving the platform or manually transferring content between tools. The document is the final deliverable of a multi- aent pipeline, not a separate task you handle after the agents are done. Ion UI ships with 20 built-in assistants out of the box, which covers a solid range of everyday tasks. But the platform is also designed to be extended, and that extension layer is the community skills marketplace, which is one of the more underrated parts of what ionui offers and one that is going to matter more as the platform grows. Skills in IONUI are agent behavior extensions. Think of them as specialized instruction sets that you attach to an agent to give it a focus capability for a specific type of work.

[00:17:43]The community marketplace is where you browse what is available, install what is relevant to your workflow, and assign skills to the agents you are running.

[00:17:50]Currently available community skills include UI design, gamedev, academic diagrams, Excel pro, financial models, coach, roleplay, and planner. With the library expanding as more contributors publish their own, the design decision that makes this particularly useful is that skills are model agnostic. The same UI design skill works whether you attach it to Claude codeex or the built-in ion CLI. You are extending the behavior layer of the agent, not modifying the model underneath it. That means you can mix and match. Attach a financial model skill to claude code for a finance heavy project. Swap it out when that project is done. Attach a different skill for the next one. The agent adapts to the work rather than you having to maintain separate agent configurations for every domain you work in. You can also create and publish your own skills. If you have built a specialized workflow around a particular type of task, a specific research methodology, a document structure you use repeatedly, a coding pattern you want every agent session to follow, you can package that as a skill, use it across all your agents, and optionally publish it to the community.

[00:18:51]This is the same model that made package managers and plug-in ecosystems so powerful in software development. Ionui skills marketplace is early, but the infrastructure is there and [music] as more developers and creators contribute to it, the compound value of the platform grows with every skill that gets added. The word co-worki's branding is doing real work and this use case is where that positioning becomes concrete.

[00:19:15]This is not about a specific feature in isolation. It is about what Ion UI enables at the level of how you actually interact with your agents while they're operating on your machine. The desktop co-work model is fundamentally different from the chat model that most AI tools are built around. And the approval layer is what makes it both powerful and safe to use on real projects. In a standard chatbased agent interaction, you send a message, the agent responds, and you decide what to do with the response. The agent is not touching anything directly.

[00:19:41]Ion UI's co-work inverts that relationship. The agents, whether that is clawed code, open claw, Hermes agent, or the built-in ion CLI, are operating directly on your computer. They are reading files from your file system, writing code to your project directory, executing scripts, calling MCP tools, browsing the web. They're doing real work on real assets, not generating text for you to act on separately. What makes this safe and practical rather than alarming is the approval layer that sits between the agent and your system. Every action that touches your machine surfaces a permission dialogue before it executes. The agent proposes the action.

[00:20:17]Write this file. Run this script. Call this tool. And you approve or reject it.

[00:20:22]The sidebar tracks pending approvals with badge counts for each active agent.

[00:20:26]So you can see at a glance what is waiting for your input across all of your running sessions. You can approve individually or handle them in batches.

[00:20:33]And the agents continue working on everything that does not require your sign off while you review what does. For agents like Clawude Code and Open Claw, which are already capable of significant autonomous action in a terminal, ion UI's approval layer adds a visual management surface that the terminal simply does not provide. You're not reading through raw terminal output trying to figure out what the agent just did. You are seeing structured, organized action requests in a UI designed for human review. The result is that you can hand genuinely complex multi-step tasks to your agents across multiple sessions simultaneously without losing situational awareness over what they're doing to your files in your system. That combination of autonomous capability and transparent oversight is what the co-work is actually about. The last use case is one that might not be the first thing you think of when you picture an AI agent platform built around tools like Claude Code and Codeex, but is more practically useful than it sounds. [music] And the reason it belongs on this list is not image generation as a standalone feature, but image generation as a native part of an agent workflow. Ion UI supports in workflow image generation directly inside the platform. With support for Gemini image models, Nano Banana, the keyword there is in workflow. You're not switching to a separate image tool, opening a new browser tab, copying a prompt somewhere else, downloading a file, and bringing it back into your project. The image generation capability is available to your agents as a tool use action inside the same session where everything else is happening. An agent can generate an image midtask the same way it would write a file or call an MCP server as one step in a larger sequence of work. The workflows this enables are genuinely multimodal in a practical sense. Hermes agent researches a topic and structures the content. Claude Code processes that content and generates the copy or the data. Ionui's image generation produces the visual assets, diagrams, UI mockups, hero images, charts rendered as graphics, and the PPT creator or word creator assembles all of it into a formatted document that is an end-to-end content pipeline running inside a single ioni session with no manual handoffs between tools, no file transfers, and no context switching. The image is not a separate deliverable you produce somewhere else. It is one output type amongst several that the agent pipeline produces as part of the same task for developers building agentic content pipelines for creators who are using agent tools to produce researchbacked visual content and for anyone running automated workflows that need visual assets as part of the output. Having image generation available at the agent level rather than as a separate tool changes how you architect those workflows. It is the difference between a pipeline that produces text and hands it off to you and a pipeline that produces the finished multimodal output from start to finish. 10 use cases and what they add up to is a platform that is genuinely trying to solve a real problem. Not just giving you another way to chat with an AI model, but giving your entire agent stack a shared environment to operate in. multi-agent orchestration, centralized tooling, continuous automation, remote access, server deployment, document generation, community skills, desktop co-work, image generation. These are not isolated features bolted onto a chat interface.

[00:23:37]They are a coherent architecture built around the idea that the value of AI agents compounds when they work together and collapses when they are isolated from each other. Ion UI is free, open- source, and available on Mac OS, Windows, and Linux. You can find on GitHub under iOSAI/ionUI.

[00:23:56]The link is in the description below. If you are running any of the agents we covered in this video, claude code, codeex, Hermes agent, open claw, open code, ioni auto detects on them the moment you install it. You do not need to reconfigure anything. Your existing setup carries over and you just gain everything on top of it. If you want to run ionui as a persistent server deployment and keep your agents running around the clock, hosting or VPS is what I recommend for hosting. reliable performance, straightforward setup, and the right hardware tier for what Ionui needs on a server. The link for that is also in the description. It helps support the channel when you use it. If you got value out of this video, the like button is the fastest way to tell the algorithm this content is worth pushing to more people. It takes 1 second and it makes a real difference for a channel like this. Subscribe if you want to stay updated as more tools in this space evolve. And drop a comment below letting me know which of these 10 use cases you're most likely to actually use. I read every comment. That is it for this one.