Last updated: 2026-07-11
Roger AI overlays live click-by-click guidance on any macOS app using Gemini vision. Free 14-day trial, no Windows support yet. See pricing and rivals.
Roger AI is a macOS screen-guidance app launched on Product Hunt in March 2026 that reads the screen at 1 frame per second using Google Gemini to overlay live click-by-click instructions on any application. It never controls the mouse or keyboard itself; it only guides. Windows support is planned but not yet shipped.
Roger AI is a screen-guidance assistant, launched on Product Hunt in March 2026, that watches whatever is on a user's display and overlays real-time instructions, highlight boxes, and tooltips showing exactly what to click or type next. Instead of replacing the user or sending them to a written tutorial, Roger sits between those two extremes: it guides while the person keeps their own hands on the mouse and keyboard, similar to screen-sharing with a patient expert standing beside them. The product runs five specialized agents orchestrated through Google Gemini's multimodal models on Google Cloud. Roger captures the screen at roughly one frame per second, detects which application is open, and reasons over raw pixels rather than the DOM, accessibility tree, or app-specific APIs. That pixel-only approach is what lets it guide a user through Excel, Figma, Photoshop, or a browser tab without a dedicated integration for each one. Users can interrupt at any point, ask a question in plain language, and resume the walkthrough where they left off, and any step that turns out wrong can be flagged for the system to correct going forward. Roger is aimed at people learning unfamiliar software fast: new hires ramping on internal tools, freelancers picking up a client's stack, or anyone stuck mid-task who would otherwise search for a tutorial. It is explicitly built as a guidance layer, not an autonomous agent: it never clicks or types on the user's behalf. The product currently ships as a native macOS app with a 14-day free trial; no paid pricing tiers have been published as of mid-2026. Windows support and a B2B SDK are on the public roadmap, alongside plans to add MCP (Model Context Protocol) support so guided workflows can connect to other agentic tools. A backend repository is public on GitHub under the TryRoger organization, and the team has said it intends to offer a bring-your-own-key option since screen captures are currently processed on Roger's own servers.
Free for a 14-day trial on macOS. No paid pricing tiers have been published as of mid-2026; the company has not disclosed post-trial cost.
Roger AI is a screen-guidance assistant that launched publicly on Product Hunt in March 2026. It watches whatever is on a user's macOS screen at roughly 1 frame per second and overlays real-time instructions: highlight boxes around the element to click, tooltips explaining what to type, and arrows pointing to the next step. It is built by a small team under the Roger AI name and runs on Google Gemini multimodal models hosted on Google Cloud. Unlike a written tutorial, Roger adapts live to whatever screen it sees, and unlike an autonomous agent, it never takes control of the mouse or keyboard itself. The goal is to help a person complete a task in an unfamiliar app while learning it, rather than outsourcing the task entirely.
Roger AI is free to use for a 14-day trial period on macOS. As of mid-2026, the company has not published any paid pricing tiers or a subscription price for continued use after the trial ends. There is no indication yet of per-seat, per-task, or usage-based billing on the public site. Teams evaluating the tool should expect to reach out directly for pricing once the trial period nears its end, since none is listed publicly. This makes budgeting for wider rollout difficult until Roger AI publishes a standard pricing page.
Roger AI's core feature is its live screen overlay, which highlights exactly where to click and what to type inside any visible application. It runs a five-agent system, including a dedicated task agent that tracks whether the overall workflow, not just the current step, has been completed. Because it reads raw pixels instead of the DOM or accessibility tree, it works across non-web software like Photoshop and Excel with zero per-app integration. Users can interrupt at any point to ask a natural-language question and then resume the walkthrough, and any incorrect step can be flagged so future guidance for that app improves.
Roger AI offers a 14-day free trial rather than a permanent free tier. During the trial, users get full access to the live screen-overlay guidance on macOS. After the 14 days, no public pricing has been announced, so it is unclear whether continued use requires payment, an upgraded plan, or is simply cut off. Users who want the tool long-term should confirm current trial terms directly on tryroger.xyz before relying on it for ongoing work, since the terms could change as the company finalizes its pricing model.
Scribe is a close alternative for teams that want to document a workflow once and share a step-by-step guide afterward, rather than getting a live in-the-moment overlay. Manus is a better fit for users who want an autonomous agent to actually execute a multi-step task rather than be guided through doing it themselves. Loom-style screen recording tools are worth considering if the goal is asynchronous training content instead of real-time, interactive guidance. Choose Roger AI specifically when the priority is learning a live application by doing it yourself with a guide alongside you.
Roger AI is best for people who need to learn unfamiliar software quickly without leaving the app to search for a tutorial, such as new hires ramping up on internal tools or freelancers picking up a client's software stack. It also suits solo users tackling complex creative or spreadsheet tools like Figma or Excel for the first time. It is not a good fit for teams that need Windows support today, since only macOS is available, or for organizations that require SOC 2 or HIPAA attestations before adopting a tool that sends screen data to a vendor's servers.
Roger AI does not currently offer a public API for developers. The company has stated that MCP (Model Context Protocol) integrations, which would let guided workflows connect to other agentic tools, are planned but not yet released as of mid-2026. A backend repository for the Windows build is publicly visible on GitHub under the TryRoger organization, though the license and scope of what is open versus proprietary is not fully documented. Developers wanting to build on Roger AI today would need to wait for the announced MCP support to ship.
Scribe focuses on capturing a workflow once, automatically turning it into a step-by-step written or visual guide that can be shared with others later, which suits documentation and training-at-scale use cases. Roger AI instead guides the user live, in real time, while they perform the task themselves on their own screen, which suits in-the-moment learning rather than producing a document. Scribe works primarily in the browser, while Roger AI is a native macOS app that can guide any visible desktop application, not just web pages. Choose Scribe to build a reusable how-to library; choose Roger AI when a person needs help completing an unfamiliar task right now.
Roger AI overlays live, step-by-step click guidance on any app by reading your screen at 1 frame per second, no integrations needed.
Roger AI