Overview
Almost every company has essential legacy software that resists any kind of automation because it lacks an API (Application Programming Interface). GUIDE brings automation to these packages anyway with a virtual keyboard and mouse, allowing the software to be used in the only way it can. This is backed up by the modern AI of your choice (such as Anthropic’s Claude).
This not only broadens the range of tasks that can be performed, it approaches the tasks in the way that is more familiar to human operators, requiring no formal automation training.
GUIDE Components
This article covers the installation of the GUIDE Windows app and its use to create and run GUIDE Agents.
For GUIDE user and account administration, see Using the GUIDE Portal.
For end-users (who can run published GUIDE Agents but not create them), see The GUIDE Web Client (TBD).
How it Works
Building a GUIDE Agent
The process starts by describing what you want GUIDE to do. This can be anything from, “Open Notepad and write ‘Hello, world!’” to multi-step spreadsheet manipulations or almost anything else that can be done on a Windows system. For more complex tasks, you can attach screen shots and other images to clarify what you mean.

GUIDE generates steps. With the help of an AI like Claude, GUIDE expands your description into steps that are more specific and AI-friendly. These steps are your GUIDE Agent.
If the steps look okay, you press the Run button to run the Agent (if not, you adjust them or start over).
GUIDE runs the Agent.
GUIDE handles the mouse and keyboard itself, outsourcing the thinking to the AI you selected.
At every step, GUIDE records a summary of its thought process and a screen shot. These will come in handy for debugging or refining the process later.

When the Agent completes, it provides statistics about resource use.
It also offers a button to “Get Quick Fix Suggestions.”
If the run was successful, the suggestions center around improved efficiency.
Otherwise, they center around fixing the problem.
If you adopt some or all of the suggestions, go back to Step 3 and run the Agent again.
When you’re satisfied, you can publish the Agent so you (and optionally members of your team) can run it at any time.
Running a GUIDE Agent

GUIDE users at the “User” level can run published Agents but can’t edit them or create new ones. Only “Builders” and “Administrators” can do that.
Those at the User level control the variables used by the Agent, such as file names or customer names. From the user’s point of view, the Agent is a simple utility.
Sign into the GUIDE desktop app or GUIDE Web app.
Choose an Agent from the list.
Press the “Run” button.
Terminology
In the context of GUIDE:
GUIDE (capitalized) is the AI enhancement package.
Development system. A Windows device running GUIDE.
An Agent (capitalized) is an Agent created and run under GUIDE. (AI agents in general are not capitalized.)
Agent system. A Windows device running GUIDE Agents. This can be a different system from the development system.
Remote device or remote system.
Implementation
GUIDE is implemented as an application that runs on Microsoft Windows. It makes API calls to cloud-based AI engines such as Anthropic’s Claude. (It currently does not support locally hosted AIs.)
GUIDE’s Agents can run locally on the GUIDE development system (the same system on which GUIDE is installed), or on a remote Agent system, where GUIDE controls the Agent via RDP (remote desktop protocol) session. When running, the Agent takes over the keyboard and mouse, and the maximum supported screen resolution is 1280x800 pixels. Running Agents on a remote Agent system is recommended except for brief periods of testing.
You create GUIDE Agents on the development system and run them either on the development system or a remote Agent system.
You can publish Agents so other members of your team can use them, and they can publish Agents that you can use.
The development system is a Windows device with Internet access (the underlying AI’s API must be accessible).
The remote Agent system, if present, is a Windows device that is accessible via Remote Desktop and has access to the Internet. The remote Agent system is optional.
Current and Recent Releases
Current Release: GUIDE 1.0.9.
GUIDE 1.10.0
Scheduled for August, 2026.
GUIDE 1.0.9
Highlights
Released on August 1, 2026.
Published Agents now have sharable links to run the Agent directly (sign in is required, of course).
GUIDE Web Client support. End-users can run Agents from the Web Client without installing GUIDE on a local Windows device. The Client uses the URL https://web.runguide.ai/customer_identifier, where “customer_identifier” is unique to your organization’s Workspot deployment.
GUIDE 1.0.8
Highlights
Released on July 14, 2026.
Support for Opus 4.8.
Initial support for Enterprise plans.
Admins can now specify which models builders can specify for their agents.
Admins can now see information at the company level in addition to the team level. This includes the overview page, graphs, and event logs.
Other Changes
The title bar of remote GUIDE windows now shows the name and execution state of the current Agent.
For security, remote RDP I/O policies have been locked down. This means that your remote GUIDE sessions no longer support audio, file transfer, printing, smartcards, local USB devices, or using the clipboard to copy data between the local and remote devices.
The "Restart remote computer before execution" checkbox has been removed.
The Agent's "Run Summary" now lists the provider and model used for the run.
Vertex AI is now called “Gemini Enterprise Agent Platform” to match following Google's name change.
GUIDE 1.0.7
Released on June 9, 2026.
The GUIDE runtime on the remote host now halts the Agent if the connection with the local GUIDE program is lost, whether through a network disconnect or a crash.
You can now specify mouse operations that involve keystrokes, such as "shift left_click," "ctrl right_click," etc.
A bug that caused runtime configurations to disappear has been fixed.
If GUIDE runs out of tokens during Plan generation, it now stops cleanly with a "token quota exceeded" message instead of retrying.
GUIDE 1.0.6
Released on May 14, 2026.
Support for Anthropic Claude Opus 4.7.
You can now select the runtime configuration (local, remote, etc.) from the Agent view.
You can now attach up to five images or PDF files in the "Build" phase. GUIDE uses these attachments to understand what you're asking for when constructing its expanded prompt but does not refer to them during runtime.
GUIDE 1.0.5
Released on April 30, 2026.
Increased execution speed.
Support for Microsoft Login (in addition to Google Login).
Users of a paid plan can now access the supported models via Vertex AI in addition to directly from Anthropic or AWS Bedrock.
The default number of iterations has been increased from 20 to 50, allowing simple Agents to run without interruption.
Many bug fixes and UI improvements.
GUIDE 1.0.4
The maximum number of iterations you can specify has been increased to 300.
Support for AWS Bedrock models in addition to Antropic. Users of a paid plan can choose their own models. Users of the free plan should contact Workspot if they need an alternate model.
Many UI improvements and bug fixes.
GUIDE 1.0.3
First production release.
Installation
Installation always involves choosing or creating a Windows system to serve as a development system, then optionally configuring a remote Agent system.
Prerequisites
Development System and Agent System
Windows.
Internet Connection (Required for AI model access and agent execution).
Simplest Installation
This uses your usual Windows system as both a development system and the Agent system.
Note: This is the simplest installation, not the recommended installation. We recommend running GUIDE agents on a remote system.
While you are running an Agent, it controls your mouse and keyboard, preventing you from using your local system. You must also reduce your monitor’s display to 1280x800 before running an Agent.
Follow the instructions in Installation Steps for the Development Machine.
Recommended Installations
If you have access to multiple Windows devices, we recommend using one of the following:
Instead of using your main Windows device for GUIDE, install GUIDE on a remotely accessible system and use it as both development system and Agent system. The installation procedure is the same as a local installation: Installation Steps for the Development Machine
Install GUIDE on either your main Windows device or a different one and use a separate Windows system as your remote Agent system.
For the development system, follow Installation Steps for the Development Machine.
For the remote Agent system, follow Remote Agent VM Setup.
Installation Steps for the Development Machine
Download GUIDE
You can download GUIDE from the GUIDE Management portal.
Download GUIDE from the GUIDE Management Portal
.png?sv=2026-02-06&spr=https&st=2026-08-18T11%3A23%3A10Z&se=2026-08-18T11%3A47%3A10Z&sr=c&sp=r&sig=DmXOsClU8BMCglQ7F48AsDPEV6UK2y747DNzlJZK6IQ%3D)
In a browser (Edge and Chrome are the browsers that have been tested), go to the GUIDE Management Portal at https://manage.runguide.ai.
Sign into your work account using Google or Microsoft authentication.
Click the “My Account” icon in the upper right corner.
Click “Download” to download the installation files, which come in a ZIP archive with a name like
GuideUserSetup_1.0.9.zip.Inside the ZIP archive are:
GuideBuildInstaller.exe,which installs GUIDE user program on your local Windows system.GuideHarnessInstaller.msi,which installs the remote GUIDE harness service on your remote Windows system.
Launch
GuideBuildInstaller(you can do this from inside the ZIP archive, without extracting it first), and install Guide as with any Windows program, as shown below.During this process, you can customize the following settings (only “Add to PATH” is enabled by default).
Installation location (“C:\Users\username\AppData\Local\Programs\Guide\”).
Name of the Start File shortcut (default is “Guide”).
Create a desktop icon
Add “Open with Guide” to the Windows Explorer file/directory context menus.
Register Guide as an editor for default file types.
Add to PATH.

Download GUIDE via a Link
It's possible that you will be given a download link to use instead of using the GUIDE management portal. If so:
In a browser, use the link to download the file.
Launch
GuideBuildInstaller.exeand install it as described above.
Configure the Development Machine
The development machine is the one on which you ran GuideBuildInstaller.exe, and is usually your local Windows system.
Signing In

Launch Guide.
Wait for “Continue with Google” and “Continue with Microsoft” to appear.
Sign in using your work account using one of these providers.
The left-hand pane will display GUIDE options.
Left-hand Navigation Pane

The left-hand navigation pane is the starting point for GUIDE activities divided into the following sections:
Agents. Lists the Agents you’ve defined, with icons for changing their settings. Your published Agents (ones shared with your team) are listed separately.
Agent Runs. Results of your recent Agent Runs.
Settings:
Global Settings: Logfile directory, model configuration, display resolution, and max. number of iterations before timing out.
Runtime Settings. Creates remote Agent configurations (the built-in local configuration isn’t configurable). See Remote Configuration.
Help & Feedback:
Getting Started. This document.
What’s New. Latest announcements and updates (on Slack).
Community. The Workspot GUIDE community (on Slack).
About. Reports the GUIDE release number.
Account: Identifies the username, email, user role (Administrator, Builder, User), and the GUIDE plan that’s in force.
Sign Out. Signs you out and returns you to the sign-in page.
The pane is quite tall and will usually require scrolling to reveal the full “Account” section. Scrolling may be required within the individual sections as well.
Configure Global Settings

Go to “Settings > Global Settings”:
Agent Folder: Where your agents are stored. (Defaults to “C:\Users\username\AppData\Local\guide\”.)
Model: The AI model used by your Agents by default.
Claude Opus 4.6 is the default and recommended model as of GUIDE 1.0.8.
You can override this on a per-Agent basis as desired.
We recommend trying different models.
Display Resolution (Important)
Must match the actual resolution of your display or RDP connection (your Agent will not run otherwise).
The maximum resolution is 1280x800 pixels.
If running on your local system, change your display resolution before running your Agent.
If running with a remote configuration, the resolution will be set automatically.
Set your desktop background to a solid color with no wallpaper (wallpaper confuses Agents).
Creating Your First Agent

Click the “+” symbol in the “Agents” section on the left-hand pane.
Create a Workflow
Fill in the Build page in the Create Agent tab:
Give the Agent a name.
Create a “Workflow,” by describing what the Agent is supposed to do. This will be parsed and turned into a more precise step-by-step version that you can review and edit as needed.
Click “Generate Steps.”
GUIDE will generate detailed steps from your Workflow and display them in a new right-hand pane.

On the right-hand pane:
Review the expanded steps. Note that only your original Workflow is editable at this point.
Click “Go To Agent.”
This takes you to the “Agent” page.
Note: Instruct your Agent to save its answer to a file. It does not do this by default, nor does it report its result in its step-by-step narrative.
Edit and Run Your Agent
.png?sv=2026-02-06&spr=https&st=2026-08-18T11%3A23%3A10Z&se=2026-08-18T11%3A47%3A10Z&sr=c&sp=r&sig=DmXOsClU8BMCglQ7F48AsDPEV6UK2y747DNzlJZK6IQ%3D)
Once in the Agent page, you can edit the steps and run the whole Agent or just selected steps.
To run the Agent, click “Select All” and then “Run.”
To run selected steps, select the step in question and then click “Run.”
Selecting random steps may have random results.
If you are asked to type in the value of a variable, do so:

How the Agent Runs

Guide now runs the Agent to completion if it can.
Its actions are broken up into “Iterations,” the amount of work it can do as a unit.
With each iteration, it uses screenshots and its instructions to decide what to do next.
It presents its thoughts in the right-hand pane, along with links to each iteration’s final screen shot.
If there was a problem with the previous iteration, the Agent figures out what went wrong and tries again. It can often solve problems on its own after no more than a few attempts.
Quick Fix Suggestions

Once The Agent is done, it gives a “Run Analysis” and an option to “Get Quick Fix Suggestions.”
If the run was successful, the Quick Fix Suggestions are to make future runs execute more efficiently.
If the run was not successful, the Quick Fix Suggestions are to make future runs successful.
To accept the fix, click “Apply to Step.”

Editing the Agent
Use the pencil icon to the left of the Agent’s name in the left-hand pane to edit the Agent or its variables.
Using Variables

The “Variables” button near the top of the Agent editing screen will show the variables defined for this Agent and allow you add more or edit or delete existing ones.
Two variables, “current_datetime” and “os_platform,” cannot be edited or deleted. These are the current date and time and the operating system architecture, version, and build.
Publishing the Agent

When editing or running the Agent, there is a “Publish” button near the top of the screen. Clicking it will publish the Agent for the rest of your team. It pops up a page first, in case you want to change some of the Agent’s fields.
Runtime Configuration

The runtime settings tell GUIDE whether to run locally or remotely, which AI model to use, the maximum number of iterations (steps) to user, and more.
To see and modify the runtime configuration:
Open the GUIDE sidebar
Under “Settings,” click “Runtime Settings.” This opens the “Runtime Settings” tab.
The installation has a “Default local Config” that can’t be edited. You can add more runtime configurations with the “+” button at the top of the page.
Default Local Config
Named Default Local Config.
Cannot be edited or deleted.
Used for local execution only.
All new agents inherit it automatically.
Adding New Runtime Configurations
To add a new runtime configuration, Click “+” in the “Runtime Settings” tab.
Local Configuration

A local configuration is shown above. It will run on the same system that GUIDE is running on.
It includes the following fields:
Configuration Name. Your choice of a name for this runtime configuration.
Description. An optional description field.
Execution Type. “Local Computer.”
Remote Configuration

When you select “Remote Computer” instead of “Local Computer,” you get a remote runtime configuration. When you select a remote runtime configuration for an Agent, it will run on the specified remote machine.
GUIDE connects to a remote Windows device using the Remote Desktop Protocol.
Make sure the remote device is accessible from the local GUIDE machine via RDP (that is, make sure you can sign in via Remote Desktop Connection).
Your RDP Credentials will be securely stored in Windows Credential Manager.
Do NOT delete credentials manually.
The configuration has these parameters:
Configuration Name. Your choice of a name for this runtime configuration.
Description. An optional description field.
Execution Type. “Local Computer.”
IP Address. This can either be the remote device’s IP address or its hostname.
Port. Currently fixed at 3389.
Username, Password. The credentials used to sign into the desired account.
Restart remote computer before execution. Disabled by default, not generally recommended.
Running Agents Remotely
Install the Remote GUIDE Harness

The GUIDE Harness is the remote GUIDE runtime that communicates with the GUIDE Builder or the GUIDE Web Client and with the selected Cloud AI provider. It is installed as a Windows service from the same ZIP archive we used for the development system. Follow these instructions to prepare the remote Windows PC or VM for the GUIDE harness:
RDP into the remote Windows device VM, using an admin account.
Follow the download procedure to acquire the latest the GUIDE release.
Extract the files from the ZIP archive. We will be using
GuideHarnessInstaller.msi.Run the installer.
Double-click the GUIDE harness Installer.
Accept the terms and click on Install.
(Optional) Navigate to Control Panel\Programs\Programs and Features and check the version details, such as the release number.

Switching Runtime at Execution Time
Once you have multiple runtime configurations, you can choose which one to run at execution time.
In “Step” view, click the dropdown next to “Run.”
Select which runtime to use for the current execution.
This does not alter the Agent’s default runtime.
GUIDE Harness Troubleshooting
GUIDE Harness Logs
You can find the GUIDE Harness logs at installation location in the “C:\ProgramData\GUIDE” folder.

Uninstall the GUIDE Harness
Uninstall normally from Control Panel.
Update the GUIDE Harness
Uninstall the old GUIDE Harness first, then install the new version.

