Raydo Agent OS

You install Raydo Desktop from the download page or with Homebrew, then connect the Chrome extension so the agent can read the page you are on. The first-run activation path walks you through Quick Entry, the browser extension, workspace mode, a runtime health check, a first result, and the choice between local use and Raydo Cloud.

Work starts from a task in the workspace, for example "research three competitors and draft a launch brief". Raydo opens the browser, reads the page context, reads the local folder and files you chose, and runs the tools the task needs. Every step shows up in a run log, so commands, browser steps and desktop actions remain inspectable after the run finishes.

Before anything is published, written, deleted or handed off, the run pauses and waits for a person to approve it. Once approved, the outputs are saved as artifacts, such as a markdown launch plan, a PDF of research notes, an image or a JSON run record, together with source notes and cost. Over time you turn tasks that work into reusable roles, skills and flows, so the next run follows a proven method instead of a blank prompt. The same work system is reachable from the desktop workspace, the raydo CLI for automation and diagnostics, and a local API for integrations.

• •

  Browser and desktop context in one run - The Chrome extension lets an agent read the live page you are on, and the desktop app reads the folders, files, screenshots and exports you choose. Both feed the same project, so a run can move from page context to local files to a finished artifact without copying material between tools.

• •

  Visible runs with approval points - Each run records the steps it took, including browser actions and commands, and shows what is running, what is blocked and what needs review. Runs pause before publishing, writing, deleting or handing off so a person can approve, correct or stop the work, and the inbox collects items waiting on you.

• •

  Saved deliverables and a delivery record - Raydo treats generated work as product material. Files, notes, assets, approvals and the run record are saved together with the task, and every artifact links back to the run that produced it, so results can be reviewed, shared, edited and reused. Because the record stays with the project, the next task can pick up from earlier outputs instead of starting from zero.

• •

  Roles, skills, flows and Raydo Hub - Work can be organised into an org with roles that define who owns a job and which tools and connections it needs. Proven methods become reusable capabilities your team can call again, the workspace includes scheduled tasks, and Raydo Hub offers a public catalogue of Work Packs and Workflow Templates with version, digest and trust checks before you install one.

• •

  Multiple entry points and adapters - Besides the desktop workspace, Raydo exposes a raydo CLI and a local API. Its GitHub README describes a multi-adapter design that includes local Codex-style and Claude Code-style adapters and gateway-based execution modes, and the changelog lists native computer use, App Studio and App Capsule delivery for generated sites and apps, video assembly, team workspaces with membership and publishing permissions, and mobile push notifications when a long-running task needs your attention.

• •

  Raydo Cloud credits for models and media - Optional Cloud plans give a single shared credit balance for chat, image and video generation across supported models, plus managed web search that returns citation-ready results without you managing a separate search key. The pricing page illustrates the Plus allowance of 20,000 credits as roughly 166 images or 250 seconds of video if spent on one type of generation, and top-ups are available at US$1 per 1,000 credits, recorded separately from subscription allowances.

Raydo is aimed at people whose work already spans the browser, local files and several AI tools. Developers can connect code, terminal work, browser checks and delivery without losing task context, and use coding agents and CLI tools against project files. Researchers can combine current web sources with local material into a traceable research artifact.

Content creators and marketers can move from an idea to research, a script or campaign brief, supporting assets and a publish-ready package in one run, with approval before anything goes out. Consultants and agencies benefit from reusing a reliable delivery method across clients while keeping each client result specific and reviewable, and team leads get a view of what is running, where judgment is needed and what is ready to hand over.

Individuals can start on the free desktop app and add a Cloud plan only when they want hosted model and media capacity, while people doing paid client work are expected to buy the Personal Commercial License. Organisations that need private deployment, permissions and auditing, device management or offline delivery can contact the team for the Enterprise option. Raydo is less suited to someone who only asks an assistant occasional questions; its own guide says an Agent OS starts to matter when AI is already part of daily work, such as preparing reports from several sources or reviewing actions before they are published.