Your records stay local
Pragmatically saves your inventory on your computer. World Director keeps your worlds and reusable model library there too.
XBAR / THE COMPANY
Local intelligence.
Personal purpose.
I build Xbar, a local-first suite for conversation, creative work, everyday inventory and games. Pragmatically is included, and XbarCade is its arcade. World Director remains a separate product under the Xbar name.
A NOTE FROM XBAR’S CREATOR
Why I built Xbar.
I built Xbar for the freedom to think, create and experiment with AI on my own computer. I wanted a companion for conversation, writing, code and creative work, with choices about the local models and tools I use.
I started Vee in Python to explore whether AI could make collaborating in Discord easier. I hoped to make a game with friends, and wanted a creative tool my wife and I could use together.
Today, Xbar brings local AI, Pragmatically and XbarCade together in one desktop suite. It also powers the separate World Director application.
The name is personal, too. I first typed Xbar as my character’s name in EverQuest. It became the name I use in every game. Counter-Strike brought me to Steam in its early days, and I’m Xbar there too.
PRIVACY & CONTROL
Your hardware.
Your choice.
With local models, Xbar’s AI processing happens on your computer. You choose the supported models and when to use connected features.
Local first, cloud second. Optional cloud models are planned for Xbar and its curated apps.
Find the right hardwareLocal dictation across Xbar and World Director
Xbar, Pragmatically and World Director use Whisper for local dictation. After the local speech engine and model are installed, your speech is transcribed on your computer. Current local speech setups use English models.
Pragmatically also offers a separate browser dictation option, which may send audio to the browser’s speech provider.
Know when you’re connected
Messages and files shared in Discord pass through Discord. Web search and online generation use their providers. Setup and model downloads may also need internet access.
HARDWARE BY APPLICATION
Start with the app.
Then choose your hardware.
You don’t need a supercomputer to get started.
Pragmatically’s basic inventory tools are the lightest starting point. Everyday Xbar text chat adds a local language model. Image, video and 3D creation need more resources—so a computer that handles chat well may still be too slow for media creation.
LIGHTEST START · INVENTORY
Pragmatically
Everyday inventory, with optional local AI for meals and projects.
- Processor
- 10th-generation Intel Core i5
or AMD Ryzen 5 4000 series - Memory
- 8 GB RAM for smaller local AIInventory-only minimum still being evaluated.
- Graphics
- Built-in graphics for everyday inventory
- Operating system
- Windows 10 or Windows 11 (64-bit)
Browse, organize and edit your inventory without running AI models. The processors are suggested starting points; the 8 GB figure is setup guidance for smaller local AI. Performance testing is ongoing.
AI, voice and phone access
The smaller text model can run on the processor without a separate graphics card. Processor speed and memory needs vary with the model.
The larger text-model option is suggested from 16 GB RAM with an NVIDIA graphics card with 8 GB of its own memory, or 24 GB RAM when using the processor alone. These are setup recommendations; performance on smaller computers is still being evaluated.
Attaching a photo is different from asking AI to identify it. Photo recognition needs a separate vision model. Local Whisper dictation runs on the processor and currently requires local AI setup.
A phone or tablet can connect through its browser on the same trusted Wi-Fi. Your host computer runs the app and any local models.
Windows 10 needs build 19041 or later. Leave storage for your chosen models, records and photos; image-, video- and 3D-generation models are not required.
SUGGESTED CHAT BASELINE*
Xbar
Local conversation, writing and brainstorming on an everyday computer.
- Processor
- 12th-generation Intel Core i7
or AMD Ryzen 7 6000 series - Memory
- 16 GB RAM
- Graphics
- Built-in Intel or AMD graphicsChat can run on the main processor; a separate graphics card is optional.
- Operating system
- 64-bit Windows
These are suggested starting points for everyday chat, not a tested minimum or a response-speed guarantee. Performance testing is ongoing.
* Media creation needs more. A computer that handles chat well can still be too slow for image, video, music or 3D creation.
Chat speed, media and local voice
My older 10th-generation Intel Core i7 computer runs local chat with Gemma 4 E4B, but some basic replies take about 30 seconds. The newer Intel and AMD suggestions above still need direct testing; they are not a claim of identical performance.
Response times depend on the exact processor, model, conversation length and the computer’s power and cooling. A newer generation alone does not guarantee quick replies.
Check the graphics, memory and storage needs of your chosen creative model. These chat specs do not establish a media-generation minimum.
You can keep chatting during a media job, but both share the computer’s memory. Larger chat models and longer conversations need more memory too.
Local Whisper dictation runs on the processor after its speech files are installed.
HEAVIER WORKLOADS · 3D WORLDS
World Director
3D creation and rendered worlds need more graphics power.
- Processor
- 14th-generation Intel Core i5
or AMD Ryzen 5 7000 series (desktop) - Memory
- 16 GB RAM
- Graphics
- NVIDIA RTX 5070Separate graphics card with 12 GB of graphics memory.
- Operating system
- Windows 11 (64-bit)
My development computer uses the Intel option. The AMD processor is a suggested alternative awaiting testing. Final requirements, including PC/VR performance, are still being tested.
Local Whisper dictation runs on the processor, separately from graphics-heavy creation.
Compare World Director targets ↓WORLD DIRECTOR ONLY
3D and PC/VR hardware targets
These tiers apply to World Director. They are not requirements for Xbar chat or Pragmatically’s everyday inventory tasks.
The Intel development baseline is my current testing computer, not a confirmed minimum for every feature. The AMD alternative and the medium and high targets have not yet been benchmarked. Final requirements and PC/VR performance are still being tested.
Compare World Director’s four development targets
Intel reference · AMD alternative
Development baseline
- Processor
- 14th-generation Intel Core i5
or AMD Ryzen 5 7000 series (desktop) - Memory
- 16 GB RAM
- Graphics
- NVIDIA RTX 507012 GB graphics memory
- Operating system
- Windows 11 (64-bit)
Proposed · not yet benchmarked
Medium target
- Processor
- AMD Ryzen 7, 9000 seriesProcessor example
- Memory
- 32 GB RAM
- Graphics
- NVIDIA RTX 508016 GB graphics memory
- Operating system
- 64-bit WindowsFinal requirements still being tested.
Proposed · not yet benchmarked
High target
- Processor
- AMD Ryzen 9, 9000 seriesProcessor example
- Memory
- 64 GB RAM
- Graphics
- NVIDIA RTX 509032 GB graphics memory
- Operating system
- 64-bit WindowsFinal requirements still being tested.
Under investigation
Ultra — experimental direction
- Processor
- High-end workstation processorConfiguration still to be decided.
- Memory
- 128 GB RAM or more
- Graphics
- RTX 5090-class or workstation graphicsMemory specified per card.
- Operating system
- 64-bit WindowsFinal requirements still being tested.
Ultra explores more demanding worlds and creation tasks. Using several graphics cards is still under investigation; faster generation and combining their memory are not confirmed capabilities. More memory or graphics cards alone do not guarantee larger worlds, better visuals or faster creation.
Computer memory (RAM) and graphics memory are separate. These targets assume a Windows computer with solid-state storage (an SSD). Storage space requirements are still being established.
About my reference computer and memory
Windows 11 · 14th-generation Intel Core i5 · 16 GB RAM · NVIDIA GeForce RTX 5070 with 12 GB graphics memory.
This desktop has run Pragmatically’s local text models and World Director’s conversation and library controls. Generating images and 3D objects needs more memory headroom than these everyday tasks.
These are planning targets, not guaranteed minimums. RAM and GPU memory are separate; performance also depends on model choice, graphics support, scene complexity and other running software.
THE ROAD AHEAD
What’s next for
the Xbar family.
Here’s where each application stands today and what I’m working toward next.
Pragmatically
Windows · Included with Xbar
Today: Inventory, household profiles, projects and recipes, with paired phone access.
Next: Refine everyday workflows, phone access and performance across more hardware through private testing.
World Director
ALPHA DEVELOPMENT
Today: Create models, build a reusable library, import compatible assets and arrange your own scenes.
Next: Improve creation, importing and scene editing. Augmented reality and sharing worlds are planned directions.
Xbar
Available for Windows
Today: The 4.2 test release adds standalone local chat, picture questions and optional parental controls, alongside its Discord connection and configured creative workflows.
Next: Refine the desktop experience and creative workflows through private testing across more setups.
Xbar is available for Windows with Pragmatically and XbarCade included. Standalone Workmate Run is available separately. World Director remains in development. Discord invites are not open yet.
Local first. Cloud second.
Your own hardware is the starting point. I plan to offer optional cloud models across Xbar, Pragmatically and World Director, so you can choose where Xbar’s AI processing happens.
This shared choice of local or cloud models is planned and is not available yet. Applications currently keep their records and conversations separate.
Other future Xbar ideas include context from folders you choose and new ways to collaborate and share resources with your permission.
A SHARED AMBITION
Mega Projects
A longer-term idea for exploring larger goals together: opt in with your interests, vote on proposals, and choose whether to contribute skills, resources or time. Participation would remain voluntary; voting would never enroll you automatically.
Concept only · no release date promisedFUTURE EXPLORATION / COMPACT WORKMATE
A companion
for the physical world.
Meet Compact Workmate: a four-wheeled robot concept with a binocular camera head, short lift and rear tray.
With Xbar running locally, it could recognize objects and suggest items for your inventory. Onboard controls would handle movement.
In the garden, it could suggest plant, fungi and lichen identifications for you to check—hands-free as you explore. A possible headset AR experience would bring that idea into your view.
Future concept · not an available feature
Interested in Xbar?
Xbar is available now; other applications and game concepts remain in development. Interested in the work or a collaboration? Email me.
operator@xbar.computer
Click to open your email app, or copy the address into a new message.
An enquiry does not subscribe you to a newsletter or guarantee application access. How I handle email.
Website privacy
Ownership and third-party materials
By Xbar · © 2026 Christopher O’Mahony.
Christopher O’Mahony reserves the rights he holds in original Xbar materials. Third-party software, models, trademarks and other materials remain subject to their respective owners’ rights and applicable licenses.
Visiting the public website
This Xbar website is operated by Christopher O’Mahony. You can browse its public pages without signing in. Cloudflare hosts and protects the site and processes request and connection information, including IP addresses. Its account dashboard provides me with aggregated traffic statistics, such as request counts, visitor or visit estimates, country-level traffic, bandwidth, errors and cache performance. These figures are not verified counts of individual people or proof that someone read a page.
I do not add advertising trackers or analytics cookies to the public website, and the separate Cloudflare Web Analytics service is not enabled. Public-page visits are not recorded in the site’s custom activity log, even if you are already signed in. Application access remains private and uses separate access controls.
Cloudflare and private sign-in services may use cookies for security and authentication, depending on the features in use. Provider processing and retention are separate from the site’s custom records. See Cloudflare’s privacy policy and Cloudflare’s cookie information. Contact operator@xbar.computer with privacy questions.
Private access and activity records
Approved private access uses sign-in and access controls. On the protected staging site, the site’s custom activity log records authenticated page requests so I can review private-site use and troubleshoot requests. It stores the verified account ID and email, time, known page route, request method, response status and a broad client category. Only I can access the activity view. Historical records from the earlier private website may remain until the cleanup described below.
This custom log does not store IP addresses, full user-agent strings, query strings, cookies, credentials or device fingerprints. The user-agent is used temporarily to derive the broad client category. Cleanup retains at most the newest 5,000 records from the preceding 30 days. It runs when activity is recorded or I review the log, so older entries may remain stored while the site is idle. Request records can include repeat or automated activity and do not prove that a person read a page.
The separate private download service is paused. Historical sign-in, request activity records and aggregate download-start counts may remain with ChatGPT Sites. OpenAI also handles platform traffic, authentication and operational information separately; see OpenAI’s privacy policy. The 30-day custom-log cleanup above does not describe every provider record, backup or download total.
When you email me
Emailing sends your address and the information you choose to include through my email provider. I use that information to reply and manage your enquiry. Contacting me does not subscribe you to a newsletter or guarantee application access; demonstrations and evaluations are arranged individually.
Please avoid sensitive personal or confidential information. For privacy questions or requests about information you have provided, contact operator@xbar.computer.