Agent Go | Agentic Copilot

Install, A to Z

Download the zip with the button above.
Unzip it to a permanent folder, for example Documents\Agent Go (macOS: ~/Documents/Agent Go). Chrome loads the extension from that folder every time it starts, so do not move or delete it afterwards. Inside you will find a folder named agent-go-extension.
Open the extensions page: paste chrome://extensions into the address bar (Edge: edge://extensions, Brave: brave://extensions).
Turn on Developer mode — the toggle at the top right of that page (in Edge it sits in the left-hand panel).
Click "Load unpacked" and pick the agent-go-extension folder — the one that contains manifest.json. Agent Go appears as a card on the page.
Pin it: click the puzzle-piece icon in the toolbar, then the pin next to Agent Go.
Open the side panel: click the Agent Go icon, or press Ctrl+Shift+G (macOS: Cmd+Shift+G). If the shortcut does nothing, another extension already uses it — assign one at chrome://extensions/shortcuts.
Sign in (before the first message): in the side panel click ⚙ Settings (top right) — or right-click the Agent Go icon → Options — then in the Account card type the same email and password as this site and click Sign in. The card switches to your email, plan and credit balance. No account yet? Create one, then sign in. If you send a message while signed out, the panel replies "Not signed in to Agent Go" with these steps and an Open Settings button.
Run your first task: go to any page, describe what you want in the side panel, press Enter. Type while it runs to steer it; ■ stops it.

Chrome shows a "Disable developer mode extensions" popup at each start while an unpacked extension is installed. Click Cancel (or the ×). It is expected for any extension installed this way, and it never affects the extension.

Before you start

Google Chrome 114 or newer on Windows or macOS (the side panel needs 114+). Edge and Brave load Chrome extensions too.
An Agentic Copilot account with credits — you sign in inside the extension with the same email and password.
A ServiceNow instance you are already logged into in a tab, if the task is about ServiceNow. Agent Go works through your own session; it never asks for your ServiceNow password.

What it costs

Included models: 0.15 credits / turn
Vendor models, where your plan allows them: 0.30 – 1.20 / turn
Bring your own key (usage plan): 0.05 / turn + your vendor's token cost

A turn is one model answer. Reading pages and ServiceNow lookups do not cost credits. The Prompt Builder is free (up to 20 prompts an hour); an image you attach to it is described first at the included-model rate. Add credits any time; the balance is on the Options page and refreshes while it is open.

Updating

Download the new zip from this page.
Empty the folder you installed from, then unzip the new files into that same folder. (A different folder would register a second, blank copy of the extension.)
On chrome://extensions click the reload arrow on the Agent Go card.

Your sign-in, settings and shortcuts are kept — they live in Chrome's storage, not in the folder. The version you are running is shown at the top of the Agent Go card on that page; this download is —.

Uninstall

chrome://extensions → Agent Go → Remove, then delete the folder you unzipped.

Advanced: local machine bridges (optional)

Out of the box the agent works inside the browser tab. Two small servers you run on your own computer let it go further: voice input transcribed locally, OS-level mouse / keyboard / screenshots for native apps, shell commands (npm, git, tests, builds) in a project folder, and text from local PDFs. They answer only on 127.0.0.1, so nothing on your network can reach them. What the agent sees through them (screenshots, command output, transcribed speech) goes to the model you use, like anything else it reads, and every desktop action or command still asks for your approval in the side panel.

Download the bridges pack version — · — KB

Needs Windows with Python 3.10+ on PATH and Agent Go 0.2.2 or newer.
Unzip anywhere permanent. INSTALL.md inside lists the pip install lines for each server.
Start desktop-server\start-desktop.bat. It prints a desktop token; paste it in Agent Go → Options → Advanced settings — local machine bridges, then switch on what you want (desktop control, Run commands) and click Test.
Start whisper-server\start-whisper.bat for local voice input (the first start downloads the model).

Keep "Enable Run commands" off when you do not need it: commands run with your own user privileges, like a terminal. The token the desktop bridge prints is the only thing that stops other programs and web pages on your computer from using it; treat it like a password. Pack checksum (SHA-256): —

If something is off

Developer mode is greyed out, or the extension disappears after a restart — the browser is managed by an organisation policy that blocks unpacked extensions. Ask your IT team to allow it, or install on a personal profile or machine.
"Manifest file is missing or unreadable" — the wrong folder was picked. Choose the agent-go-extension folder that contains manifest.json, not the zip file or the folder above it.
Sign-in says the account has no access yet — write to info@nowidevops.com and the account is enabled for Agent Go.
The side panel is blank, or a step never finishes — on chrome://extensions click the reload arrow on the Agent Go card and open the panel again. If a page tool still misbehaves, reload that page too (Ctrl+F5).
"Insufficient credits" — the balance ran out mid-task. Add credits and send the task again; nothing was charged for the turn that did not run.
Voice buttons use Chrome's speech service — normal. The 🎤 dictation and 🎧 listener run on the browser's built-in speech recognition.

Privacy

Page content and screenshots you send are transmitted to the Agentic Copilot service and the model provider that answers. Do not send confidential data. The extension never stores your password; a bring-your-own-key, if you configure one, stays on your machine only.