Let it see and act

Most of your work already lives somewhere: in folders, in a cloud drive, in your inbox, in a supplier's website that wants a password. Player One can reach all of it, and it can install what it needs to finish the job.

Your files

Say "here are five supplier invoices, pull the totals into a table", and drop the PDFs onto the chat.

Player One reads them, writes the table up and opens it in the preview pane beside the conversation, so you watch it being made rather than waiting for an attachment. Attach files with the paperclip, drag and drop, or paste; a dropped folder keeps its structure.

The file lands in a workspace: the named place your agents keep documents, data and the output of a job. You find it later under Manage > Files (/workspaces), where you can browse, search, preview, upload and download.

See Workspaces and files and The preview pane.

Your cloud drives

Say "sync the Clients folder from our Dropbox into a workspace and keep it up to date".

You ask for the drive to be connected, sign in to it in your own browser when the link arrives, then say which folders come in and which way they travel: two way, download only (cloud into the workspace), or upload only (workspace out to the cloud). Google Drive, Dropbox, OneDrive, Box and SharePoint connect by signing in; WebDAV, S3 and SFTP take a short credential form. Only the storage services your installation has been set up with are offered.

Each synced folder reports for itself on the workspace's Cloud sync tab: when it last ran, live progress, and buttons to sync now, run a full resync or detach. Detaching stops future syncs and deletes nothing.

Two limits before you rely on it: a code repository's hidden Git data never syncs, because a two way sync would corrupt it, so a synced copy is a working copy rather than a clone; and if more than half the files on one side vanish, the sync stops itself and changes nothing until you decide what happened.

Steps in Connect a cloud drive.

Your mail and your diary

Say "connect my Google account, then tell me what I have on Thursday and draft a reply to the Acme thread".

There is no settings page for this one. You ask in the chat and your assistant walks you through it one Google page at a time, naming the heading and button to look for. You choose the services you need (Calendar, Gmail, Drive, Docs, Sheets, Slides and Tasks are covered) and whether Player One may only read them or also write. Someone with administrator rights on your Google account completes the Google side, once for each Player One installation, and the credentials go into your vault through a secure form rather than an ordinary chat message.

Email and the other channels you talk to it on are connected under Admin > Messaging (/admin/channels): see Channels and communication.

Websites it can use with your logins

Say "log into the supplier portal and download this month's invoices".

Player One opens a real browser, gets as far as the sign-in page on its own, then puts a secure browser window beside the chat and asks you to take over. You click inside it, type your password, deal with the verification code, and press Hand back. The agent picks up in the same browser, on the page you left it on.

Your password is typed into the website in that window: it never passes through the AI, and you never have to send it in chat.

The browser remembers. Each login is kept in a named profile, so the next time that portal comes up it is already signed in. You can also say "show me what you are doing", which gives a view-only window with a Take control button for when you want the keyboard.

Three practical notes. A control session normally lasts thirty minutes, though the login it leaves behind does not expire with it. Your own password manager and passkeys will not usually appear inside that window, so pick the site's QR code, authenticator or text-message route. A file downloaded in there starts on the Player One machine before copying to whatever you are viewing from.

Worked example: Log in for it once, let it work from then on. The browser describes the browser itself: watching it live, profiles, downloads and limits.

The same browser works without you: screenshot a page, capture it as a PDF, pull a table off it, or record a short video of a site.

The services you already pay for

Say "open a GitHub issue for that bug", "add a Trello card", or "file this in Notion".

Player One ships with skills for a long list of services, and each uses your account rather than a shared one. You hand over the key or token once through a secure prompt, it goes into the vault, and your agents then use it by reference without ever seeing the value. Browse what is installed under Admin > Skills (/admin/skills); Skills explains what a skill is. Vault and sandboxing explains how the secrets are held, and Stay in control covers what an agent may do with them.

Tools it installs for itself

Say "you will need a converter for this, install one".

Your installation is a small computer of its own, so your agents can install command line tools, Python packages and npm globals, and those survive a restart. Something that must keep running, such as a watcher or a sync client, becomes a background service that starts on boot and comes back after a crash. It can also connect to an MCP server.

It runs without administrator rights, so a tool that demands them will not work, and anything installed outside the persistent area disappears at the next restart. Whether an agent asks you before it installs or runs something depends on its permission mode, Conservative, Most things or Everything, which Stay in control explains.

Your own computer

Say "what's on my screen?", "fill this in on my machine", or "run the backup script on the office PC".

Install Sidekick on a computer and sign in once, and your agents can work on that machine: look at the screen and the open windows, click and type in a program, run commands in a shell that stays open between steps, and read or write the folders you share. You choose how much, per computer, from Off through Observe only and Change things to Full access. A new connection starts at Observe only, and your choice is kept across restarts. When an agent wants the real mouse and keyboard, a prompt asks you first, and Take control in the tray stops it until you hand back.

It is a power-user feature with no sandbox: an agent you allow to change things can change them. The same app on Windows and on a Mac. Sidekick also comes as a browser extension for Chrome, Brave and Edge, so an agent can work in the web pages you allow, in your own browser with your own sign-ins.

Steps in Connect your own computer; Sidekick describes the app and the four choices.

Machines, servers and networks

Say "the office server is not responding, get on it and find out why", or "set up the new backup server and connect it to the NAS".

Your agents are as at home on a machine as in a document. Give one access to a computer or server, usually over SSH with the key or password kept in the vault, and it does the work a technician would: reads the logs, finds what is failing, repairs it, installs and configures what is missing, and reports back in plain English. The same goes for the network. It keeps a router or firewall configured, checks what is on the network and why something cannot reach something else, and connects a new server into the rest. Microelectronics are within reach too: a board, a sensor or a controller it can speak to over a serial or network link is one more thing it can read, program and monitor.

A Windows machine on your own network works the same way over SSH, once that is switched on at the Windows end. Whatever it finds and changes goes into memory, so the next time that machine misbehaves the history is already there, and a scheduled check can watch it for you from then on.