Raydo Agent OS

Most AI chat tools end with an answer in a window. Raydo Agent OS is built around the work that comes after, where the source is a web page, the material is a folder on your laptop and the result has to be a file someone checks before it goes anywhere. The homepage example is a launch package: research three competitors, read local files, draft a brief, then stop and wait for approval before saving the final copy.

That stop is the main idea. A run moves from the browser through your desktop to an approval point and then to a saved artifact, and each step is logged, from the page opened and the text extracted to the local folder read and the tool run paused. The outputs land together as a delivery record, so a later task can start from launch_plan.md and research_notes.pdf instead of from nothing.

Raydo is a desktop app for Apple Silicon Macs and Windows, with a Chrome extension for page context and a raydo command-line tool for automation. Inside, work sits in projects with an inbox, chat, runs and scheduled tasks. You can define roles, which hold the context and tools a job needs, save proven methods as reusable skills and string steps into flows. The site has guides for developers, researchers, content creators, marketers, consultants, agencies and team leads, each with example tasks.

The app also generates images, audio and video. Paid plans run on shared credits: Plus is $20 a month for 20,000 credits, Pro $100 for 100,000 and Max $200 for 200,000, with annual billing about 30% cheaper and top-ups at $1 per 1,000 credits. The pricing page gives sample rates, such as GPT Image 2 at 120 credits an image and Grok Imagine at 80 credits a second of 720p video, and notes that real usage depends on the model and current prices. A separate personal commercial license covers commercial use of the app itself.

Raydo is run by Rolitt Inc., a Delaware company. Its privacy policy describes the app as local-first but says some features may send prompts, files, workflow metadata or outputs to the model provider, cloud service or integration you pick. The default Google Workspace connection asks for read-only access to Docs and Sheets and reads only the files you select as inputs. Because agents can read local folders and browse on your behalf, it's worth starting with a test project and checking what each role is allowed to touch.

It suits people whose AI work already spans tabs, files and several tools, and who want a reviewable record of what an agent did. Someone who just needs quick answers will find a chat app simpler.