theagi.company

AGI, Inc. is a company building private, on-device action AI for phones and computers. Its core products are AGI-0, a small "action model," and an agent that sees the screen and acts across apps on Android — without APIs or integrations. The company positions this as a privacy-first alternative to cloud-dependent assistants, and it serves two audiences: individuals who want hands-free control of their devices, and developers who want to build action agents on its platform.

What AGI-0 actually is

AGI-0 is described as a small action model that lives on the device rather than in the cloud. Two properties follow from that design:

• It learns locally. The site states AGI-0 learns your preferences and improves over time, and that it remembers what you like.
• It works without app integrations. Instead of connecting to services through APIs, the agent sees the screen and acts across any app on Android — the same way a person would tap through an interface.

The company reports AGI-0 scoring 100% on AndroidWorld (116/116 tasks), a benchmark for Android task automation. That figure comes from AGI, Inc. itself and reflects a specific benchmark, not general real-world reliability.

What the agent can do

The site frames the product around hands-free task completion across everyday app categories. Its own examples show the intended scope:

These examples share a pattern: a natural-language goal that spans multiple apps and requires reading context (messages, schedules, routes) before acting. That is the capability the "action model" label refers to — not conversation, but execution.

The site also states the agent works offline, on wearables, and for privacy-sensitive tasks. Those claims are the company's own; no independent testing is provided on the page.

Who it's for

Individuals who want to offload busywork on their phone — the pitch is explicitly about reclaiming time ("Focus on your life, not your screen"). The privacy angle matters most for people uncomfortable with a cloud assistant reading their screen.

Developers who want to ship an action agent quickly. The company says you can "ship a working action agent in less than a day" using its platform, and references MCP (Model Context Protocol) in its site navigation, suggesting a protocol-based integration path for builders.

What to verify before committing

The page does not list pricing, subscription terms, or login requirements, so cost and access conditions are unknown from this source alone. Before signing up, check:

1. Device support — Android is named explicitly; the site mentions phones and computers, but confirm your specific hardware, including wearables.
2. Access model — the site has "Get access" and "Book a call" calls to action, which suggests gated or waitlist-based onboarding rather than open sign-up.
3. Benchmark context — the 116/116 AndroidWorld result is a controlled benchmark; ask how the agent handles apps and flows outside that set.
4. Data handling — "private by design" and on-device processing are stated, but the specifics of what leaves the device (if anything) are not detailed on the page.

If your goal is hands-free Android automation with local processing, AGI, Inc. is directly relevant. If you need cross-platform support beyond Android, documented pricing, or third-party validation of the benchmark claims, those answers aren't on the site and are worth requesting directly.

Yes—AGI, Inc. explicitly invites developers to build on its platform, and its site claims you can "ship a working action agent in less than a day." That claim is worth testing against your own use case, because the platform's design choices (an on-device action model, no API integrations, screen-level control) shape both what you can build and what you need to verify before committing.

What the platform is built on

AGI, Inc. builds AGI-0, described as a small action model that lives on the device, plus an agent that sees the screen and acts across any app on Android. Two properties follow from that architecture and matter most to developers:

• No APIs or integrations required. The agent operates through the screen rather than through app APIs, so in principle it can act inside apps that expose no developer interface.
• Private by design and on-device. The model runs locally, learns user preferences over time, and the company states it works offline, on wearables, and for privacy-sensitive tasks.

The site also reports that AGI-0 scores 100% on AndroidWorld (116/116 tasks). Treat that as a vendor-reported benchmark: it tells you the model handles a defined task suite, not that it will handle your specific workflows.

Evaluate these dimensions before you build

The access row deserves emphasis: the site offers "Get access" and "Book a call" rather than self-serve signup details, and it publishes no pricing. Do not assume free or open access—confirm eligibility and terms directly.

A practical evaluation path

1. Define one narrow task you would want an agent to complete—for example, ordering a specific set of grocery items in a shopping app, or booking a ride conditionally when a train is delayed (both are task types the site itself illustrates).
2. Request access and book a call through the site's "Get access" and "Book a call" entry points, and ask specifically about supported devices, app coverage, and privacy handling during onboarding.
3. Build the smallest agent the platform allows and time it, to test the "less than a day" claim against your team's actual setup effort.
4. Test on your real apps and devices, not just demo scenarios—screen-based agents are sensitive to UI variation, so check the apps your users rely on.
5. Confirm the privacy story end to end, since on-device operation is the platform's central selling point and the main reason to choose it over API-based automation.

When this platform fits—and when it doesn't

It is a reasonable fit if you need an agent that operates inside third-party Android apps you don't control, and if keeping action data on the device is a priority for your users or your compliance posture. It is a weaker fit if your workflows already have clean APIs (direct integration will be more predictable), if you need platforms beyond those AGI, Inc. supports, or if you require published pricing and self-serve onboarding before you can start.

The decisive step is the call: the site's claims about speed, coverage, and privacy are specific enough to test, but the access and commercial details you need for a build decision are not on the page.

AGI, Inc.'s hands-free agent can handle multi-step tasks that normally require you to open several apps and tap through them yourself — ordering groceries on Instacart, comparing routes and booking a rideshare, finding a long-term activity that fits your schedule, and locating a playlist in your messages and playing it on Spotify. The key condition is that the agent works by seeing the screen and acting across apps on Android, rather than through app APIs or integrations, so it operates the same apps you already use.

How the agent approaches a task

AGI, Inc. builds AGI-0, described on its site as a small action model, plus an agent that "sees the screen and acts across any app on Android." The company says AGI-0 lives on the device, learns your preferences over time, and works offline, on wearables, and for privacy-sensitive tasks. That on-device design is also why the site frames the product as "private by design."

Practically, this means you state a goal in natural language and the agent carries out the steps inside your apps. The examples below come from the prompts shown on the company's site.

Task examples the site demonstrates

Two things stand out in these examples. First, several tasks are conditional or long-running — "if the train is delayed" and "for the next 6 months" — so the agent is meant to handle decisions and time windows, not just single taps. Second, the tasks cross app boundaries: Messages to Spotify, calendar to a booking site.

What to check before relying on it

• Platform: The action agent is described for Android phones and devices. If you use another mobile platform, the cross-app screen control described here may not apply.
• App coverage: The site's claim is "any app," with the examples above as illustrations. For a task you care about, confirm the specific app is one the agent can operate in your setup.
• Access: The site shows "Get access" and "Book a call" entry points rather than a self-serve download, so availability may be limited or gated. The page evidence does not state pricing or whether access is free.
• Benchmark context: The site reports AGI-0 at "100% on AndroidWorld (116/116 tasks)." That is the company's own reported figure on a benchmark of Android tasks, not a guarantee for your particular apps or account state.

Who this fits

This kind of agent suits people who want to delegate routine, multi-app chores — restocking groceries, coordinating a trip, scheduling recurring activities, or digging up something a friend sent — without switching between apps manually. If your work stays inside one app or you need guaranteed behavior on a specific service, test that exact task first; the value of a screen-acting agent depends on the apps and conditions you actually use.

AGI, Inc.'s privacy model rests on one architectural choice: AGI-0, its action model, lives on the device rather than in the cloud. Because the agent sees the screen and acts across apps directly on the phone, it doesn't need APIs or third-party integrations to reach your data — and the company frames the result as "private by design." This matters most if you want an agent that can operate across sensitive apps without shipping your screen contents to a remote service.

What "on-device" actually changes

The company describes AGI-0 as a "small action model" that "lives on the device." That single fact drives the rest of the privacy story:

• No API layer to expose. The agent works by seeing the screen and acting across any app on Android, rather than by connecting to services through APIs or integrations. Fewer integration endpoints means fewer places where credentials or data are handed to a third party.
• Screen content stays local. Since the model runs where the screen is, the observation-and-action loop happens on the phone itself. The site's own framing is that data stays on the user's device.
• It works offline. AGI, Inc. states AGI-0 "works offline, on wearables, and privacy-sensitive tasks" — an offline capability is only possible if the core action loop doesn't depend on a remote server.

A concrete example

Take the site's own demo prompt: "Find the playlist Jake sent me in Messages and play it on Spotify." A cloud-integrated agent would typically need read access to your messaging app and a Spotify connection. AGI-0 instead reads the screen, finds the message, and drives the Spotify app the way you would. The message content and your account aren't passed through an integration — the action happens on the device.

Where privacy and capability meet

The design isn't only about withholding data; it's also what enables certain tasks. The company positions AGI-0 for "privacy-sensitive tasks" and for wearables, where a cloud round-trip would be impractical. It also says the model "learns your preferences and gets better over time" while living on the device — so personalization is described as happening locally rather than by training on a server-side profile.

What the page does not establish

The site is a product and marketing page, so treat the privacy claims as the company's stated design intent, not an audited guarantee. It does not publish a technical whitepaper on the page, detail what telemetry (if any) leaves the device, or describe encryption or data-retention practices. If your decision depends on those specifics — for example, deploying it in a regulated environment — ask AGI, Inc. directly through the "Book a call" or "Get access" routes on the site before relying on the on-device claim.

Who this design suits

Choose this approach if your priority is keeping screen-level activity on the phone and you want an agent that reaches apps without you wiring up integrations. Look elsewhere — or get answers in writing first — if you need documented compliance controls, third-party security audits, or clarity on any data that may still leave the device. The privacy-by-design pitch is coherent and specific about mechanism; the open questions are about verification, not architecture.

AGI-0 is a small action model built by AGI, Inc. that runs on the device and operates Android apps by seeing the screen and acting on it, rather than connecting through APIs or integrations. It is designed for hands-free task completion across the apps already on your phone, and because it lives on the device it can work offline and learn your preferences over time. It suits users who want an agent to handle multi-step phone tasks privately, without setting up app integrations.

How AGI-0 works

AGI-0 takes a spoken or written instruction, reads what is on the screen, and performs the taps and actions needed to complete the task inside whatever app is relevant. The company describes it as an agent that "sees the screen and acts across any app on Android," with no APIs or integrations required. Two consequences follow from that design:

• No setup per app. Because it works from the screen rather than an API, it is not limited to apps that offer integrations.
• On-device processing. AGI-0 lives on the phone, which is the basis for its privacy and offline claims.

AGI, Inc. also reports that AGI-0 scored 100% on AndroidWorld, completing 116 of 116 tasks. That figure comes from the company itself and refers to a benchmark, not a guarantee for every real-world task.

What it can do

The site groups example tasks into four areas, each showing the kind of multi-step request the agent is meant to handle:

These examples share a pattern: the task spans more than one app or requires a decision (compare routes, check a schedule, locate a message), and the agent is expected to carry it through rather than just open an app.

Privacy, offline use, and learning

AGI, Inc. states that AGI-0 runs on the device, learns your preferences and improves over time, and works offline, on wearables, and for privacy-sensitive tasks. In practice this means tasks that don't need a network connection can still be attempted, and the agent's behavior is meant to adapt to how you like things done. The site does not publish details on exactly what data stays on the device versus what is shared, so treat the privacy framing as the company's stated design goal rather than a fully documented specification.

Who it is for

AGI-0 is aimed at Android users who want an agent to take over routine phone work — ordering, booking, scheduling, finding and playing media — so they spend less time tapping through screens. It is also relevant to developers: AGI, Inc. offers a platform for building action agents, which the company says can get a working agent shipped in less than a day. If your priority is an agent that works inside apps without integration work and keeps processing on the device, AGI-0 is built around that; if you need documented enterprise data-handling guarantees, the public materials don't yet provide them.

Access is handled through the company's "Get access" and "Book a call" routes on its site; pricing and availability details are not specified in the available materials.