ASL AI Hub — Source library
Public documentation snapshot: 2026-09-08
One request. The right AI for every step.
Describe the outcome; AI Hub picks the right model for each step, on your accounts, with a receipt.
One request. The right AI for every step.
Built for: A solo consultant or developer paying for two or three AI providers who wants one place to send work and one honest bill · A small team that shares AI accounts and needs budgets, approvals, and a record of who ran what · A business owner who has to show where a request went, which model handled it, and what it cost · Someone who already pays for a ChatGPT or Claude plan and wants it to do approved work from a phone or desktop without handing anyone the password
ASL AI Hub exists because of this.
You pay for two or three AI providers. Every few weeks a different model is best at something, prices move, one provider has a bad day, and the model you trusted quietly gets worse at the one thing you used it for. So you pick by habit. When the first answer is weak you paste the same prompt into a second tab. You never really know what any of it cost. If you run a team it is worse: keys get shared in chat, nobody can say who ran what, and the bill arrives as one number with no story behind it. The tools that promise to fix this mostly put themselves in the middle. An aggregator takes your prompt, forwards it through its own account, and sells you a markup. Now a third company has your data and your provider relationship, and you still choose the model yourself from a dropdown. Meanwhile the ChatGPT or Claude plan you already pay for every month sits idle for real work, because there is no safe way to point it at a task from your phone. What you actually want is simple to say. Describe the job. Let something competent decide which model should do each part. Keep your keys and your prompts out of anyone else's hands. Get a receipt you could show your accountant or an auditor. That is the job AI Hub does.
ASL AI Hub exists because of this.
Direct to each provider: there is no aggregator or reseller in the middle, so nobody else holds your prompts or your provider relationship
ASL AI Hub exists because of this.
The task is the primary object: classification happens inside AI Hub before any model is chosen, and the request is never sent to a provider to pick a provider
ASL AI Hub exists because of this.
Hard capability, policy, and budget gates run before any scoring; accuracy is the first gate and price is the last
ASL AI Hub exists because of this.
Preferred and usable routes are always shown separately, so a missing key never silently downgrades your work
ASL AI Hub exists because of this.
Cost is labelled Exact, Estimate, or Unknown; a missing cost stays unknown and cannot be coerced into zero
ASL AI Hub exists because of this.
Your existing ChatGPT or Claude plan can do five named kinds of approved work through the official client, with no password, cookie, or token ever touching AI Hub
ASL AI Hub exists because of this.
Every run returns a signed receipt, and outcome feedback binds to that receipt rather than to free text
How ASL AI Hub works, start to finish.
ASL AI Hub is a control plane for AI work. You sign in, connect the AI accounts you already have, and describe what done looks like. AI Hub classifies the request into one of fourteen task types inside AI Hub, before any provider sees it, so your request is never sent to a model just to choose a model. It then removes every model that cannot do the job: wrong modality, too little context, a provider your policy forbids, a lifecycle you don't accept, unknown pricing, or a cost over your cap. Only then does it rank what is left on quality evidence, speed, price, remaining budget, and reliability. You see the recommended route in plain language, the runner-up alternatives, and an honest cost label before anything runs. Compound work becomes a plan. Ask for research plus analysis plus a written deliverable and AI Hub splits it into steps, chooses a policy-compliant model for each, shows you the whole graph, and waits for your one-time approval. Independent steps can run in parallel, up to a limit your policy sets. A managed workflow can pause, resume, and be audited later, and its record never contains your prompts, evidence, or model outputs. Five routing modes (Quality, Balanced, Economy, Fast, Conserve) change what the ranking favours; they never loosen a security, privacy, or capability requirement. Your keys stay yours. Provider API keys live in an encrypted broker that makes the calls on your behalf and never returns the key to your browser or phone. You can hold several accounts per provider, each with its own routing priority, monthly budget, allowed modes, and health state, and you can revoke any of them with a recorded reason. If you would rather use the ChatGPT or Claude plan you already pay for, a small desktop companion pairs your official Codex or Claude Code sign-in to your AI Hub workspace with a one-time code. AI Hub never sees that password, cookie, or token; the companion only attests that the official client is signed in and that you are present. Every run ends in a signed receipt. Usage is reconciled from what the provider actually reported, a cost is labelled Exact, Estimate, or Unknown, and AI Hub never shows zero as proof that work was free. Spend caps refuse a request before the provider is called, at organization, project, and per-workflow level. Organization administrators set policy with teeth: which providers, tasks, and modes are allowed, budgets, retention, and whether high-risk work needs a named approver. Projects tighten those rules further for a client or a team and can never loosen them, because every project setting is intersected or capped against the organization ceiling instead of replacing it. Separate roles keep the person who writes policy apart from the one who audits it and the one who approves. AI Hub also learns. After a run you rate the outcome and mark rework or a failure reason; that feedback binds to the receipt, not to free text, and becomes evidence for that task family. An administrator can switch on a bounded challenger that tries one already-eligible alternative on a small slice of work under its own spend cap; it is off by default and never calls two providers just to compare answers. Beyond text, the same workspace handles structured data extraction, code, image generation, image and document inputs, and cited web research, with uploaded and generated files scanned, kept encrypted in your workspace, and fetched through short-lived links. For developers, AI Hub also publishes a durable, versioned working memory to AI coding tools over the Model Context Protocol, so a session's context outlives the session. All of this runs as real apps on one account: an installable web app, a genuine native Android app, and a Windows companion, with device sessions you can name and revoke, one-time recovery codes, and a security history that never contains a secret.
Everything in the current release.
Each of these is built and working today. Nothing on this list is a roadmap item.
Describe the outcome, not the model
Type what you want finished. AI Hub sorts it into one of 14 task types, from writing and long-document summary to code, code review, complex reasoning, data extraction, image analysis, image and video generation, research, voice, translation, and embeddings. Classification happens inside AI Hub, so your request is not sent to a provider just to choose a provider. When the classifier is unsure it asks you to confirm instead of guessing.
Capability first, then your priorities
Models that cannot do the job are removed before any scoring: wrong modality, not enough context, a provider your policy forbids, unknown pricing, or a cost over your cap. Only the survivors are ranked, and five modes (Quality, Balanced, Economy, Fast, Conserve) decide what that ranking favours. A cheap wrong answer is still expensive, so accuracy is the first gate. The route AI Hub would pick with every account connected is always shown separately from the best route you can run right now, so a missing key never rewrites the truth and an unconnected model is never presented as runnable.
Compound work becomes a plan you approve once
A request with several deliverables is split into a graph of steps, each with its own capability needs, model, and cost estimate; independent steps can run in parallel up to a limit your policy sets. You see the full graph and approve it one time before any provider is called. Managed workflows survive interruptions, can be resumed, and can be audited later, without the record ever holding your prompts, evidence, or model outputs.
Your provider keys never leave the vault
API keys are stored encrypted and used by a broker that makes the call for you; the key is never returned to your browser or phone. Hold several accounts per provider, each with a routing priority, a monthly budget, allowed modes, and an environment label. A circuit breaker pulls a failing account out of service and lets it back in gradually, and you can revoke any credential at any time with a recorded reason. Deleting your account revokes every credential AI Hub held for you first, and the deletion is refused if that cleanup fails.
Use the ChatGPT or Claude plan you already pay for
A small Windows companion pairs your official Codex or Claude Code sign-in to your workspace with a one-time code that expires in ten minutes. Access is a two-minute lease renewed by a heartbeat only while the companion stays open; close it and the lease fails closed. Exactly five task types can run this way: write or rewrite, complex reasoning, translate, generate code, and review code. AI Hub never handles the consumer password, cookies, or tokens, and your subscription allowance is never dressed up as API dollars.
Nine providers, connected directly
OpenAI, Anthropic, Google, Moonshot/Kimi, DeepSeek, Black Forest Labs, Runway, xAI, and Roboflow, 26 models in all, every entry carrying its own official source link and verification date. There is no aggregator in the middle: your request goes straight to the provider you connected. If the catalog's evidence ages past the freshness limit you set, routing stops before a model is chosen rather than deciding on facts that have gone out of date.
A signed receipt for every run, with honest cost labels
Each execution returns a signed receipt with the provider's own usage figures. Cost is labelled Exact when the provider reported it, Estimate when it was calculated before the run, and Unknown when the provider gave nothing back. A missing cost stays unreconciled and cannot be coerced into a numeric zero, so a run is never made to look free. The usage view deliberately excludes prompts, outputs, and credentials.
Spend caps that refuse before spending
Set a monthly organization budget, a per-workflow ceiling, and per-account budgets, with a warning percentage and a hard block at the limit. A request that would exceed a cap is refused before the provider is called. Failed calls never book their estimate as real spend. Long outputs are continued in bounded segments with the worst-case cost reserved first, and an incomplete result is refused rather than passed off as done.
Policy with teeth, for the organization and each project
Five roles (Member, Organization Admin, Operator, Auditor, Workflow Approver) keep writing policy, auditing it, and approving work in separate hands. Policy covers which providers, tasks, and modes are allowed, budgets, retention, and high-risk approval, and it is enforced by AI Hub itself so no client can loosen it. Projects bind approved accounts, connectors, budgets, privacy defaults, and members for one client or team; every project value is intersected or capped against the organization ceiling, so tightening is the only direction available. Every revision needs a reason and lands in an audit trail.
Learn from what actually worked
After a run, rate it one to five, mark it accepted or not, flag rework, and pick a reason from a fixed list. That feedback binds to the execution receipt, stores no free text, and becomes routing evidence for that task family. An administrator can also switch on a bounded challenger that sends a small slice of work to one already-eligible alternative under its own spend cap; it is off unless someone turns it on, and it never calls two providers just to compare their answers.
Privacy you can check yourself
Prompts and outputs are not stored in workflow, usage, or feedback records. Stored working context is encrypted at rest with a separate key per workspace, and deleting a workspace destroys its key so the deletion reaches backups too. An export carries a checksum you can verify with one command. The audit trail is hash-chained and anchored at every start, on every deletion, and on a six-hour sweep, so edits, mid-chain deletions, and truncation are detectable.
One account on web, Android, and Windows
Sign in once with the same username and password on the installable web app, the genuine native Android app, and the Windows companion. Name your device sessions, sign out any of them remotely or all others at once, rotate your password, and keep ten one-time recovery codes. Your security history lists events only; passwords, tokens, prompts, and outputs never appear in it. On Android, screenshots of the app are blocked.
Beyond text: images, files, code, and cited research
Generate images, proven live through OpenAI, Google, and Runway. Send images and documents as inputs, get validated structured output, and run web research that comes back with citations. Uploaded and generated files are scanned, stored encrypted in your workspace, and fetched through download links that expire in 15 minutes, bound to one workspace and one file. Embeddings, short video, realtime voice, and a sandboxed coding agent are built and bounded in the same workspace but have not yet been proven on a live provider run, and we say so rather than counting them as shipped.
Shared memory for your AI tools
AI Hub keeps a durable, versioned working context per scope and publishes it to AI coding tools over the Model Context Protocol, so a session's memory outlives the session. Your workspace is derived from your sign-in, never something you pick by hand. A write over the size cap is refused rather than silently truncated, and a write that would shrink a large document more than fivefold is held back as a likely wrong scope until you confirm it. That endpoint also answers a quick provider-preference question by task kind; the full capability-gated routing described above runs in the AI Hub app, not on this endpoint.
Numbers we can stand behind.
Every figure below comes from the product's own release record or test suite, not from a marketing estimate.
Numbers we can stand behind.
Policy can only get tighter as it flows from organization to project to request, because each level is intersected against the one above it
Numbers we can stand behind.
Prompts and outputs stay out of workflow, usage, and feedback records; deletion destroys the workspace key; exports carry a verifiable checksum
Numbers we can stand behind.
A genuine native Android app on the same account as the web app, not a wrapped web page, with revocable device sessions
Numbers we can stand behind.
What is proven and what is not is written down with dates; limits travel with every claim
Surfaces and status.
Status as of 2026-09-02. The platform is deployed and running: an immutable release was deployed and accepted on 2026-09-02, health checks returned healthy when re-checked that day, and the authenticated Model Context Protocol endpoint and customer API routes answer publicly. Real provider work has been executed with signed receipts across nine directly connected providers, and on 2026-09-01 another ASL product's AI features were cut over to run through AI Hub, producing live verified receipts, citations and reconciled cost. The signed Android app is distributed directly and works against that live API. What is not there yet: the customer web app is not served publicly at all (the public root redirects to the main ASL site and carries a no-index header, and browser acceptance to date ran against a local copy), the signed Android release is not on Google Play, and subscription checkout runs only against a sandbox with no live charges. Access is arranged directly with ASL rather than by open sign-up, and product notes dated 2026-09-01 state the customer surface stays private until an explicit launch decision.
Worth more together.
Products on this platform share one sign-in, one support queue, and one engineering standard. These pair naturally with ASL AI Hub.
Outside Access
Its AI features already run through AI Hub's sponsored lane: classification, cited research, document drafting, and image generation, with the same signed receipts and a spend ceiling. A working, dated example of AI Hub behind another product.
ASL Vault
Teams that like how AI Hub keeps provider keys sealed and never returns them to a client usually want the same discipline for the rest of their secrets.
Policy Lens
Answers that cite their sources and stay silent when the evidence is not there; a natural companion for teams who chose AI Hub for its refusal to guess.
ASL Scan
For teams that must show their AI use is governed, an independent look at the rest of their surface completes the picture.
Recent progress.
This product ships often. The most recent verified changes, newest first.
Recent progress.
2026-08-21 per-user workspaces, encryption at rest, self-service deletion that destroys the workspace key, and the tamper-evident audit chain went live; first real provider run in production.
Recent progress.
2026-08-23 encrypted provider credentials, the authenticated public app route, sandbox subscription checkout with the full billing lifecycle exercised, and a production-signed Android 0.14.8 verified on a physical Pixel 8.
Recent progress.
2026-08-24 to 08-26: routing engine v2, a live shadow evaluation across every credential-backed language model, and the advanced capability runtime (images, documents, structured output, embeddings, sandboxed coding agent) built and mounted, with image generation proven end to end through OpenAI.
Recent progress.
2026-08-29 DeepSeek, Black Forest Labs, and Runway added as direct providers; routing engine V3 and classifier V5; a guard for platform-funded usage.
Recent progress.
2026-08-30 team seats with 72-hour one-time invitations, plan sync with seat-overage handling, a revenue view, a fail-closed release gate, a USD 2-capped customer-style acceptance that built four real deliverables through three providers for USD 0.0399176 of exact reconciled spend, Android 0.14.10 production-signed, and bounded continuation for long outputs.
Recent progress.
2026-08-31 private compute-node contract live, Gemini Tier 1 validated, xAI Grok and Roboflow added and live-proven, and a customer interface refresh.
Recent progress.
2026-09-01 browser acceptance fixes against a local copy, a single signed-out entry gate, a Help Center, another ASL product's AI cut over to run through AI Hub's sponsored lane with live receipts and citations, a four-provider image comparison that proved image generation on OpenAI, Google, and Runway, and Android 0.14.11 built but not signed.
Recent progress.
2026-09-02 an immutable release deployed and accepted with a live image run; then, at source level and not yet deployed, exact image cost reconciliation from official provider rates, a controlled existing-customer acceptance runner, Android 0.14.12 metadata, and 535 automated checks passing.
Recent progress.
Pricing for ASL AI Hub is quoted after a short conversation about your situation, because the right scope differs from one team to the next. There is no charge for that conversation.
What does it cost?
Pricing is not published yet. Plans with seat limits for an individual, a small team, and a policy-and-audit-minded organization are being finalized, and checkout currently runs against a sandbox. Provider usage on accounts you connect is billed by the provider at the provider's rates; AI Hub adds no markup to your own accounts. If you want in now, ask for early access and we will talk about terms directly.
We already call OpenAI directly. Why put anything in between?
Because which model is best changes every few weeks, and every app that hardcodes a model name has to be edited and redeployed to keep up. AI Hub moves that decision to one place, adds budgets and approvals, moves to the next eligible provider on a transient error, and hands you a signed receipt per run. It is not a proxy that forwards your prompt through someone else's account: your request goes straight to the provider you connected. And you can still pin an exact model when you want to.
We already run a gateway or router. What is different?
Most routers are inference resellers: your prompt passes through their account and you pick a model from a list. AI Hub connects to each provider directly, classifies the task before it chooses a model, removes anything that fails your policy or budget before scoring, and separates the route it would prefer from the route you can actually run. It also lets a ChatGPT or Claude plan do five named kinds of approved work through the official client, which a gateway cannot do.
Where do our prompts and data go?
To the provider you connected, directly, and to nowhere else. AI Hub does not store prompts or outputs in workflow, usage, or feedback records. Provider keys are encrypted and are never returned to your browser or phone. Stored working context is encrypted with a key per workspace; deleting the workspace destroys that key so backups cannot bring it back. An export carries a checksum you can verify yourself.
What happens if we stop paying?
A lapsed subscription moves through a grace period and then suspends the workspace rather than deleting it. Your provider accounts stay yours throughout, you can revoke any key AI Hub holds at any time, and you can export or delete your workspace yourself. There is no lock-in on your keys or your prompts, because AI Hub never held the prompts to begin with.
Is it actually ready, or is this a demo?
It is deployed and running, and it is in early access. Real work has been routed through nine directly connected providers with signed receipts, four customer-style deliverables were built through three providers for four cents of exact reconciled spend, and 535 automated checks pass. What is not done: the web app is not publicly served at all, the Android app is not on Google Play, checkout runs against a sandbox, and off-site audit anchoring is still ahead. We list those limits on this page because we would rather you know.
Can I use my ChatGPT Plus or Claude Pro plan instead of API keys?
Yes, for five named task types. A small Windows companion pairs your official Codex or Claude Code sign-in to your workspace with a one-time code; access lasts only while the companion is open and you are present. Write or rewrite, complex reasoning, translate, generate code, and review code can run this way. AI Hub never sees your provider password, cookies, or tokens, and it will not pretend your subscription is a pool of API credit.
Can we run it on our own hardware?
Not today. AI Hub is hosted by ASL, and every provider call is made directly to the provider you chose. The groundwork for registering private compute alongside AI Hub exists but is not a customer offering yet. If private inference on your own machines matters to you, tell us; it shapes what we build next.
How do we get access, and what will we actually be able to use?
Access is arranged directly with ASL. In practice that means an account on the hosted platform, the signed Android app installed directly, and the authenticated integration endpoint for AI coding tools. The browser app is not publicly served yet, so if a web interface is the thing you need, say so up front and we will tell you honestly where that sits rather than sell you a date.
What should I know before I rely on it?
We would rather you hear this from us than discover it later. As of 2026-09-02:
What should I know before I rely on it?
Early access only: access is arranged directly with ASL rather than by open sign-up.
What should I know before I rely on it?
The customer web app is not publicly served. Its public route is deliberately closed and the public root redirects to the main ASL site, so today the practical way in is the Android app plus the authenticated API. Browser acceptance so far has been run against a local copy, not the hosted platform.
What should I know before I rely on it?
The signed Android app is installed directly for early-access customers; it is not on Google Play, and there is no iPhone app.
What should I know before I rely on it?
The Android interface is English only. A Spanish resource file exists in the source, but the app never reads from it, so nothing in the interface is actually translated. The web app is English only as well.
What should I know before I rely on it?
The Windows companion is a working build; it is unsigned, not in the Microsoft Store, and not yet distributable as a direct download.
What should I know before I rely on it?
Subscription checkout runs against a sandbox only; no live charges are taken and plan prices are not final. Provider usage on your own accounts is billed by the provider, not by AI Hub.
What should I know before I rely on it?
The consumer-plan lane needs the companion open, the official client signed in, and you present. It covers exactly five task types (write or rewrite, complex reasoning, translate, generate code, review code), with no web tools, files, images, embeddings, or voice, and no silent fallback to an API key.
What should I know before I rely on it?
The audit trail is tamper-evident, not tamper-proof: edits, mid-chain deletions, and truncation are detected, but because anchors currently live with the data, a complete wipe of one workspace's audit state would leave no evidence there. Off-site anchoring is the next hardening step and is not done.
What should I know before I rely on it?
Black Forest Labs image generation and DeepSeek text depend on a funded account on your side; the platform's own accounts for those two providers are unfunded, so they are not offered on the platform-funded lane. Roboflow has been proven only against a public project; the private-project path is unverified and non-public content stays blocked.
What should I know before I rely on it?
Some providers do not return an exact per-image charge with the response; AI Hub labels that cost pending rather than inventing a number.
One request. The right AI for every step.
Prefer email? contact@autosecurelogin.com