Skip to guide
ASL AI Hub

ASL AI Hub desktop help center

Support: support@autosecurelogin.com. Include the app version, exact error, first and most recent occurrence with timezone, and the last successful stage. Never send provider tokens, fleet keys, private signing keys, complete journals or customer inputs/results without reviewing what is needed and authorized.

Start here: AI Hub Companion in plain words

In plain words: AI Hub Companion helps you use more than one computer for real work. You describe a job in ordinary words, choose which of your computers should do it, check the plan and approve it, and the result comes back to a folder you choose. It can also connect AI services you already use. Nothing happens by surprise: opening the app starts no job, connection or purchase, and every step that sends work asks you first.

How to use this guide: each numbered section starts with a short In plain words explanation, followed by the exact technical details for whoever sets things up. Section 18 explains the words the app uses, section 17 answers common questions, and section 19 says where to start when something does not work.

A good first week:

  1. Connect to your fleet (section 3) and run the small practice job in section 11.
  2. When that works, plan your own work (section 4), approve it and check the delivered result (section 6).
  3. If a computer in your fleet has its own image program set up, try making a picture from a prompt (section 15).

1. AI providers and fleet work are separate

In plain words: AI Hub Companion does two separate jobs. AI providers connects AI services you already use. Fleet workflows lets you plan work, describe it in ordinary words, and send it to computers you control, such as a powerful workstation, then get the results back on your laptop. Setting up one does not switch on the other.

AI Hub Companion is a native Windows application with two independent workspaces. AI providers connects the supported provider/pairing workflow to your chosen AI Hub workspace. Fleet workflows plans approved work on machines you control, through your chosen rBroker coordinator. Neither connection authorizes the other.

AI Hub Companion is free to download. Customers pay separately for the ASL-hosted AI Hub service; that subscription is not a purchase of this desktop app. The service account provides plan selection, Stripe-hosted checkout and billing management when available. Customer-operated fleet hosting and third-party provider accounts remain separate. The Companion does not contain an embedded card-entry form or silently subscribe you when you pair or join a fleet.

You can use AI Hub without PC Steward or Master Suite. Remote execution needs a compatible, enrolled resource node such as rBroker; merely installing AI Hub on a laptop does not create a GPU server on another PC. rBroker also works without AI Hub for local integrated GPU clients, CPU jobs and fleet administration.

PC Steward contributes diagnostics and readable resource observations, not a second scheduler. Master Suite embeds these native capabilities in one separately purchased application with its own protected profiles. It does not require three other installed programs, and it does not silently import their credentials.

Opening a tab does not purchase AI usage, join a fleet, approve a workflow or install an always-running Windows service. Review each explicit operation.

2. Connect a provider workspace

In plain words: This is where you link an AI service you already have an account with. You sign in through that service's own pairing steps; never paste passwords or browser cookies into the app. What the AI service costs is between you and that service.

Use the AI providers workspace to enter the trusted workspace address and follow its supported sign-in/pairing process. Select the intended provider and review the current connection state before requesting work. A signed-in browser or official AI client is not automatically an authorized connection to this app.

When pairing an official desktop client, use the supported one-time pairing flow. Do not paste browser cookies, a Windows password or an unrelated session's token into a connection field. Provider accounts, workspace permission and local client availability are separate checks; an expired pairing is not permission to reuse another person's account.

Usage limits, API charges, subscriptions and model availability belong to the provider and account you configured. Downloading the free Companion or subscribing to the ASL service does not include unlimited third-party AI usage. Cost labels must distinguish a measured charge, an estimate and an unknown amount. A connection failure is not proof that a call was free or that an earlier operation did not run.

3. Connect a fleet across different networks

In plain words: Your computers find each other through one shared meeting point on a secure web address (the fleet coordinator). Each computer contacts that meeting point itself, so you do not need to open ports on your router. To send work you need a dispatcher key from whoever runs the fleet; it is different from a key that only lets you look.

The fleet needs a coordinator reachable by all participating machines over trusted HTTPS. Its hostname is configurable; ASL hosting is not mandatory. A Linux server or suitably configured Windows server can host it. The server operator owns certificate, backup, storage and availability responsibilities.

In Fleet workflows, configure the intended fleet and named dispatcher credential. Use the credential for this role, not a viewer key or node private key. Save the profile in this application's protected local storage. Do not put the private profile in a shared project folder or source-control repository.

On each worker computer, use rBroker or Suite's resource workspace to join the same fleet. Join fleet submits public enrollment information; manual JSON export/import remains available. The node's private signing key stays there. Then explicitly approve the dispatcher, permitted handlers, resource ceilings and consent expiry. Fleet membership by itself does not permit execution.

Nodes contact the coordinator outbound. Do not expose their loopback GPU broker or CPU-worker ports publicly or disable a firewall as a setup shortcut. Viewing resources is not an allocation, and a connected but unapproved node cannot run your job merely because its screen is green.

4. Create a plan that explains the work

In plain words: A workflow is a plan of one or more steps written in ordinary words: what the step is called, why it is done and what it should produce. For each step you choose the computer, the kind of work, how much of that computer it may use, the input and where the result should go. You save the plan, read it, and only then approve it.

Give the overall workflow and every step a title, project, purpose, stage and expected result. Prefer “Prepare forest scene for episode 1” to a session ID. Keep IDs for precise correlation, not as the only explanation. Do not include private prompts, credentials or sensitive file paths just to make a label longer.

Choose the actual target computer and one of its registered handlers. Set worker count, system-RAM requirement and any GPU-memory requirement. Select the exact input and final destination. Add dependencies so a later step cannot start before its required input is verified. Save the plan, read it, and approve that exact saved version before advancing it. A draft or validated form is not submitted work.

The current reviewed handlers cover program bundles, bounded JSON work, hashing, the supported shape-animation/media path and pictures from a prompt (section 15). A handler is a registered local capability, not an arbitrary remote shell. Media execution requires an explicitly reviewed local encoder and supported hardware. The app does not automatically install creative tools, models or third-party licenses on every enrolled computer.

A useful pattern is a CPU/RAM-rich workstation preparing scene data, a GPU workstation performing supported encoding, and the everyday laptop verifying and receiving the output. This illustrates scheduling and delivery; it is not a promise that an unrestricted cartoon prompt becomes a production-quality film or that every existing creative application has a registered adapter.

Choose Run a program with its data for general CPU computation. Select a ZIP containing a trusted Python program or Windows executable and its data. The compressed bundle limit is 16 GiB, the expanded-input limit is 32 GiB, and declared returned files total at most 16 GiB, plus bounded logs. Actual free space on the sending computer, worker and destination can impose a lower limit.

The small JSON description is separate from the file bytes; it is not a one-megabyte file limit.

Enter the relative entry point, one argument per line, and the relative files to return. Review the bundle size and SHA-256, target computer, RAM, time limit and Maximum returned data in MiB (1–16384) before approving. The return limit defaults to 2048 MiB; increase it deliberately when your expected result is larger. One bundle uses one worker; independent steps may target different computers. Program results are ZIP archives and cannot be used as JSON dependencies.

General program bundles are CPU-only; selecting a GPU does not make an arbitrary program or Blender project a distributed GPU render.

Each receiving computer separately allows program work and grants the named dispatcher access. Old JSON consent and upgrades do not enable it. Programs run with the worker Windows account's access, so send only trusted code. CPU/RAM limits and process-tree shutdown are not an untrusted-code security sandbox. Bundled Python is included; other executable dependencies belong in the ZIP. Keep free space for input chunks, verified copies, expanded files and results.

Interrupted transfers resume on the original workflow; a bounded client operation may pause after five minutes, and Run / reconcile one pass continues it without reposting accepted work.

5. Approval, advancement and actual resources

In plain words: Approving a plan lets it start; it does not take over anyone's computer. The receiving computer still checks that you are allowed, that it can do this kind of work and that it has room, and only then reserves processor workers, main memory and, for graphics work, graphics memory for your step.

Advancing an approved workflow offers eligible steps with their saved identities. The selected node validates scope, permission, handler availability and current ownership, then obtains local resource admission before executing. A claim is ownership of an attempt, not a GPU lease or a RAM grant by itself.

CPU worker slots and declared system RAM are admitted by the canonical parallel-worker authority. Supported GPU work also obtains its own GPU RAM lease. These are separate pools. System RAM is not graphics memory; a machine with spare GPU bytes may still lack CPU slots, system RAM or an approved handler. Declared RAM admission is scheduling accounting, not a physical-page pin or an OS hard limit on unrelated programs.

The node has the final decision. AI Hub does not bypass a stale capacity check, borrow another task's lease or extend an expired grant. Authentication rejection is not a temporary grace period. Read the original task/attempt and stopped state instead of creating duplicate work to make a conflict disappear.

The local Windows allowance is one configured GPU budget item. OMEN acceptance uses one 5,120 MiB allowance; this is not imposed as a capacity promise on every customer card. Status observations and driver allocations are labeled separately from committed leases. Unknown memory is never treated as free.

6. Input files, results and delivery

In plain words: Files travel directly between the two computers, encrypted, like a private courier; the fleet server only handles the paperwork. If the connection breaks, the transfer continues where it stopped. A result only counts as delivered once it has been checked and saved in the folder you chose, and an existing file is never overwritten.

Approved content transfers use separate fleet/job/artifact-scoped authorization. The receiver verifies the completed content before publishing a final file to the chosen destination. An existing unrelated file is not silently replaced.

Follow the displayed stage: queued, admitted, executing, stopped, transferring and delivered mean different things. A completed render with an unfinished transfer is not a delivered result. A subsequent dependency uses the verified predecessor output, not an unverified filename from an earlier attempt.

If the network drops, reopen the original workflow and resume. Confirmed chunks are reused; missing chunks are sent. A lost upload acknowledgement is reconciled with the same transfer. Do not render the job again just because a response was lost or the laptop was temporarily offline.

Files travel directly between the two authorized computers over an encrypted peer connection. The coordinator retains permissions, job state, manifests and connection messages; it does not store or relay new input or result bytes. Keep both applications open and both computers online until delivery is verified. A service-hosted CPU node has its own separately approved lifecycle; closing the sending application can still prevent file delivery.

Keep enough local storage for verified chunks, the bundle snapshot, expanded inputs and results. A 2 GiB ZIP can require considerably more than 2 GiB of free worker space. Files use 256 KiB resumable chunks and smaller encrypted network frames. Previously accepted server files remain readable for recovery; new transfers never fall back to uploading content to the coordinator.

Different networks can work when the computers establish a direct connection. Restrictive NAT or firewall rules can prevent this even when HTTPS fleet status works. A connection-discovery service exchanges network addresses, not file contents. Leave the original workflow and partial data intact, restore permitted connectivity and use Run / reconcile one pass. Do not expose local GPU/CPU ports, disable protection, or create a duplicate job to bypass a blocked connection.

7. Recover, cancel and archive

In plain words: If something is interrupted, reopen the same workflow; the app works out what already happened, so the same job is never run twice by accident. Cancelling asks the other computer to stop; it is not instant. When everything is delivered you can archive the workflow to tidy the list; your files are kept.

Reopen the same saved profile after interruption. Preserve the workflow ID, approved plan hash, node fingerprint, attempt IDs and receipts. Unknown means the outcome is not confirmed; it does not mean the work failed and can be repeated. An exact-offer retry reconciles the same saved request rather than inventing a second job. Do not rotate a node key or erase a journal to bypass a warning.

Cancellation requests a stop from the owning node. It is not an immediate release of someone else's resources. Claimed ownership remains until stopped evidence is confirmed. A stale screen, disappeared PID or 404 after a service restart cannot by itself prove the original work never ran.

Archive completed workflows explicitly after required destinations are verified. Archive history preserves readable plans and receipts in protected storage and retains replay protection; it does not delete your delivered files. Pending or uncertain work cannot be concealed by archiving it. History is bounded and paged; use the archive view instead of interpreting only the recent page as all history.

Damaged or missing initialized workflow storage fails closed. Preserve it and the original identity for recovery. Do not delete the private directory, restore only half a profile or open new-schema state with an older writer. See the packaged/repository FLEET-WORKFLOW-ARCHIVES guide for operator-level details.

8. Closing the app and always-running nodes

In plain words: Closing AI Hub does not stop work already running on another computer; reopen the workflow later to check it. A computer that should keep working when nobody is signed in needs rBroker's optional background service, installed with administrator approval.

AI Hub plans and advances workflows; rBroker or Suite resource hosting owns the node executors and local engines. Closing AI Hub is not a command to kill remote accepted work. Reopen the same profile to reconcile the original plan. Do not assume a closed monitoring window means the remote job stopped.

For a node that should work without a signed-in desktop, install the optional rBroker Windows service with explicit administrator approval. Its LocalService identity and CPU-only execution consent are separate from the desktop profile. Installation is not automatic when you buy or open AI Hub. Sleep, power loss and network loss still affect availability. Drain accepted work before a service upgrade; never restart a shared broker merely to test this screen.

Headless Linux examples require their own compatible runtime and validation. Apple platform scaffolding is not tested macOS hardware support. Windows 10 laptop acceptance remains a separate physical test, not something proved by a minimum-version manifest.

9. Troubleshooting in order

In plain words: Work down this list from the top. Most problems are a wrong address or key, a missing permission on the receiving computer, or a connection that dropped and only needs the same workflow continued.

10. Verify your setup and get support

In plain words: Try one small, harmless job first and check that the result arrives. When you contact support, say what you expected, what happened and when, and include version numbers; never send passwords, keys or private files.

Start with non-sensitive sample data and one modest CPU job. Watch its readable title on the intended node, confirm the final output digest and destination, and verify its worker/RAM commitment releases afterward. Then test your supported GPU handler with its own small lease. Confirm an interrupted transfer resumes under the original job rather than executing twice.

Keep source/build tests, installed-package behavior, hardware acceptance, Store submission and public availability separate. A new help guide is not an automatic upgrade to an older installation. Review the version shown by each app and server.

Email support@autosecurelogin.com with the versions, readable task description, exact failing action and state, first/last occurrence with timezone, and last successful step. Review exports locally first. Do not email private configuration, full payloads or credentials. No support response-time guarantee is implied.

Content safety and reporting: the Help center's Report inappropriate content button opens a draft addressed to support@autosecurelogin.com. Nothing is sent automatically. It reads or attaches no task, prompt, result, identity, credential, private file or log. Enter a brief concern, app version and time yourself; do not attach harmful media or confidential material. If Windows has no email app, use the displayed support address yourself. Opening a draft is not a submitted report.

There is no public feed or promise of automatic moderation.

Use only material and providers you are authorized to use. Do not use ASL tools for child sexual exploitation, non-consensual intimate content, threats, hateful abuse, fraud, malware or other illegal harm. Do not request or distribute pornography or graphic real-world violence. Respect privacy, intellectual property, provider policies and fleet-owner rules. Review generated material before sharing it. Contact the private fleet's owner as well as ASL for concerns in a customer-operated deployment.

Stop automatic advancement and use supported cancellation/consent controls; preserve unresolved ownership and recovery state. ASL reviews concerns involving its products or operated services and can restrict affected access where it has control. The customer server owner controls its retained copies; a report cannot silently remove data on someone else's server.

11. First complete CPU workflow

In plain words: A first practice run: join a computer as a worker, allow it to take work, connect AI Hub, describe a tiny job, approve it and watch the result arrive in your folder. Doing this once teaches the whole process.

  1. On the worker computer, open rBroker Connect this machine, enter the fleet address and join key, then choose Join as a CPU worker. This starts or connects to the reviewed local worker; it does not approve programs, folders or GPU work.
  2. In the worker's Fleet page, allow the required handler with modest worker/RAM limits and an expiry. The fleet manager separately grants the same named dispatcher the required scope. Enable automatic fleet execution only after both choices are confirmed. A service identity needs its own setup and consent.
  3. In Fleet workflows → Connect, enter that coordinator's HTTPS address and separate dispatcher key. Choose Connect and protect profile and check the intended computer's identity and capabilities.
  4. Under Build a workflow, describe the workflow and step, select Prepare or copy JSON data, enter harmless sample text, request one worker with a modest system-RAM amount, and choose a new final output folder and filename.
  5. Select Add step to plan, then Save private draft for review. Under Saved work, choose Read selected plan and progress. Check the computer, input, handler, resources and destination before Approve reviewed plan.
  6. Choose Run / reconcile one pass. Automatic progress is a separate opt-in for the selected workflow. Follow the stage until the verified result is delivered and the node releases its resources. Keep both apps open through file delivery.
  7. Reopen the same protected profile to inspect or reconcile the saved work. Choose Archive verified finished workflow only after completion and any selected delivery are confirmed. Archiving keeps the receipt and does not erase your output files.

Use a small non-sensitive job first. Approval covers the saved plan, not later edits. An enrolled computer with no execution consent can report its health while refusing every workflow.

12. NVIDIA, AMD and Blender projects

In plain words: Besides processor work you can send two fixed kinds of graphics-card work, short shape animations for NVIDIA or AMD cards, and pictures from a prompt (section 15). None of them turns a Blender project into a render farm.

The two GPU choices are Encode NVIDIA shape animation and Encode AMD shape animation. These are bounded shape-animation encoders. On the receiving node, open Host & setup → Local GPU media setup, choose the matching NVIDIA (NVENC) or AMD (AMF) backend and the intended existing GPU authority, then Choose and validate local media tools. Compatible FFmpeg and the graphics driver must already be available. An explicitly selected GPU Session Broker 1.3.2 can remain the authority.

Enable GPU media for that app-owned worker, and separately permit the matching handler in both node consent and dispatcher permission. An already-running worker does not gain a changed media configuration; drain owned work before deliberately restarting that owned worker. Never stop an unrelated broker. The GPU attempt needs its own GPU lease as well as CPU-worker and system-RAM admission. A backend mismatch or unavailable encoder fails visibly; it does not quietly switch to CPU.

A .blend file can be included as data in a trusted program ZIP, but transferring the file does not install Blender, create an adapter, choose render settings or split the scene across GPUs. The general program route remains CPU-only. Include compatible dependencies and your own reviewed CPU program when using that route. The built-in GPU choices do not render arbitrary Blender scenes or accept an unrestricted cartoon prompt.

13. Diagnose a transfer without repeating the computation

In plain words: When a result does not arrive, these checks tell you which part is stuck, so you fix that part without running the whole job again.

14. Version and platform checklist

In plain words: Every computer involved needs a recent enough version for files to move directly between them. Check what is actually installed on each one; reading this guide does not update anything.

The current direct-transfer workflow requires coordinator 0.16.0 or newer, compatible rBroker 1.4.5 or newer nodes, and AI Hub Companion 0.14.10 or newer or Master Suite 0.4.14 or newer. Those minimums are unchanged for existing work.

Picture steps (section 15) additionally need fleet coordinator 0.17.1 or newer and rBroker 1.4.7 or newer on the computer that makes the pictures; the sending side needs an AI Hub Companion or Master Suite version that lists Make pictures from a prompt. Older status-only connections may still work without supporting current execution or file transfers.

Compare the versions actually installed on every participating computer. Check the selected handler and driver on the actual worker, and keep its configured GPU authority unchanged. The Windows desktop and included peer runtime are the qualified path here; older headless Linux examples do not establish current direct-transfer hardware acceptance on Linux or macOS. A package's Windows minimum version is separate from physical testing on your Windows edition.

15. Make pictures from a prompt

In plain words: You can ask one of your fleet's computers to draw pictures from a written description, called a prompt, for example: a lighthouse on a rocky point at dusk, long exposure photograph. That computer makes 1 to 4 pictures with the image program its owner installed, and they come back to a folder you choose as a ZIP file holding the pictures and a recipe (the settings used to make them).

You only describe the picture; you do not need to know anything about the image program.

Before you start

Add a picture step, step by step

  1. Open Fleet workflows, connect your profile and describe the workflow (section 4).
  2. Under Build a workflow, choose the computer, then the work type Make pictures from a prompt (on that computer's own image program). The picture fields appear and the resources are filled in.
  3. Give the step a title, purpose, stage and expected result, as for any step.
  4. Type the description in Describe the picture (3 to 1,000 characters). Line breaks are turned into spaces.
  5. Optionally type what to avoid in Leave out (up to 300 characters), for example: text, watermark.
  6. Choose the Shape: square (1024 x 1024), portrait (832 x 1216) or landscape (1216 x 832), and How many pictures (1 to 4).
  7. Choose the final output folder; the file is called pictures.zip unless you change it, and an existing file is never replaced.
  8. Choose Add step to plan, then Save private draft for review. Under Saved work, read the plan: it shows the exact prompt, shape, count and leave-out text. Approve it, then choose Run / reconcile one pass or turn on Automatic progress.
  9. When the step shows delivered, open pictures.zip in your folder. Windows opens a ZIP like a folder; the pictures are 1.png, 2.png and so on, and recipe.json records how they were made.

What can stop a picture

Technical details

The step uses the rBroker work type media.sdxl-image with inline data of exactly prompt, avoid, shape and count, resources of one worker, 18,432 MiB system RAM and 14,336 MiB GPU, and one item named result. The app checks the request with the same rules as rBroker's vendored media-image.mjs before saving; the saved plan is checked again by rBroker's offer validation, by the coordinator and by the receiving computer.

The receiving computer reserves the GPU lease only while its image program runs, and its result is a stored ZIP that rBroker checks (entries, CRCs, PNG digests, sizes and the recipe's prompt and shape) before it is offered. AI Hub then receives it directly over the encrypted peer connection and verifies it before saving it to your folder. A picture step cannot be the JSON input of a later step.

The job's readable title is visible to the fleet coordinator; the full prompt travels only to the computer that makes the pictures.

16. When the connection drops

In plain words: If your internet connection or the fleet server is unavailable for a moment, nothing you approved is lost. The workflow keeps every saved request, and when the connection is back you continue the same workflow; it never sends a second copy of a job.

Choose Run / reconcile one pass to continue. Automatic progress turns itself off when something needs your attention, so turn it on again once the connection is back. Transfers continue from the last confirmed piece. On the receiving side, rBroker 1.4.7 and later no longer stop automatic work after a short outage: they keep their saved work and try again by themselves, after one minute and then longer waits up to fifteen minutes while the outage lasts.

Technical detail: a lost reply leaves the step in its saved state (for example offer-unconfirmed). The next pass reconciles the original job ID and the exact saved request; it never creates a replacement job. A transport failure is retryable; 401/403, ownership conflicts and invalid records are not, and they stop automatic progress until you review them.

17. Frequently asked questions

Does opening AI Hub start or buy anything?

No. Opening it starts no job, connection, subscription or service. Each of those is a separate choice.

Can AI Hub use another computer without its owner's permission?

No. The receiving computer's owner chooses which kinds of work it accepts and how much of it may be used, and the fleet's manager chooses which dispatcher may send. Both must agree before anything runs.

Do my files go through ASL's servers?

No. Files travel directly between your two computers, encrypted. The fleet server only carries permissions, short descriptions and connection messages.

Does AI Hub include an AI model or image generator?

No. AI providers are services you already use. Pictures from a prompt are made by an image program the receiving computer's owner installed (section 15).

What happens if my laptop sleeps during a workflow?

Saved work is kept. Wake it, reopen the same workflow and continue with Run / reconcile one pass; nothing is sent twice.

Why does a computer show online but never take my step?

Online only means it reports its status. It also needs its owner's permission for that kind of work, your dispatcher's permission from the fleet manager, enough free workers and memory, and an unexpired permission (sections 3 and 9).

18. Words you will see

In plain words: The words the app uses, explained without jargon.

19. Something is not working: where to start

In plain words: Find the line that matches what you see and do the first check before anything else.

Guide updated October 15, 2026. Generated from the same guide embedded in the native product.

Help / Ayuda