Security that starts with a separate machine
ProductiveBot runs on its own dedicated Mac, apart from the computer you use every day. Your files and agent memory live on that machine, macOS security sits underneath everything, and built-in safeguards help your agent check what it's asked to do.
Separate by design
Most DIY AI agents get installed on the laptop you already use, the one with your personal email, saved passwords, and banking. Anything on that computer is within the agent's reach.
ProductiveBot gives your agent its own computer instead. It works with the accounts and files you choose to give it, and your everyday computer stays out of the picture.
- Your agent works only with the accounts and files you give it
- Your personal email, passwords, and photos stay on your own computer
- Reset or retire the agent's Mac without touching your main one
Five layers, from the hardware up
Each layer adds protection on top of the one below it, so no single layer has to be perfect. ProductiveBot works with open agent engines like OpenClaw and Hermes Agent, and most of these layers sit underneath the engine, so they protect you whichever one you run.
Dedicated hardware
A Mac with one job: running your agents. Because it's separate from your everyday computer, a problem on one doesn't spill onto the other.
Built on macOS
Everything runs on the Mac's built-in security architecture. macOS asks before an app can reach things like your calendar, contacts, or files, so your agent has to be granted access first.
Local files and memory
Your files, passwords, and agent memory stay on your Mac, not on our servers. TotalRecall, the memory system, indexes and searches them right on the machine.
ProductiveBot safeguards
SecuritySuite, PromptGuard, and SkillGuard come installed. Your agent uses them to check commands, incoming content, and new skills for risk.
Permissions you control
You decide which accounts and apps your agent can use, and who is allowed to message it. Cloud models and outside services connect only when you set them up.
Three safeguards, installed and ready
They come set up on your machine, so you don't have to find, vet, or configure security tools yourself.
SecuritySuite
Checks commands, links, and files for risk
Your agent can run a command, link, or file path past SecuritySuite before acting on it. It checks against known-dangerous patterns, like deleting system files or reaching addresses it shouldn't, and logs anything it flags on your Mac.
PromptGuard
Helps detect risky instructions
Emails, web pages, and documents can hide instructions meant to trick an AI agent. PromptGuard checks incoming content for these injection attempts, rates how risky they look, and logs what it flags.
SkillGuard
Helps assess skills before they're installed
Skills add new abilities to your agent, and community-made ones can hide harmful code. SkillGuard scans a skill for signs of credential theft, code injection, or data being sent where it shouldn't. ProductiveBot's own built-in skills are scanned with it during setup.
Prompt injection is an unsolved problem across the AI industry, and no tool catches everything. That is why these checks sit on top of a separate machine and permissions you control, not in place of them.
What stays on your Mac, and what connects out
We'd rather tell you up front.
Stays on your Mac
Your files and documents that you give your agent
Agent memory, including TotalRecall's index and your conversation history
Passwords and app logins your agent uses
Requests to a local AI model, if you run one
Connects out, when you use it
Requests to cloud AI models, like ChatGPT or Claude, when you use them
The apps you connect, like email, calendar, or Slack
Support chats with Scout, when you ask for help
Your purchase email, once, to activate your license
Software updates and Skill Store downloads
ProductiveBot is not air-gapped. Your agents use the internet to reach the AI models and apps you connect. Your files and memory are not stored on our servers. Read our guide to where your AI data goes, or see how ProductiveBot is built.
Help without holding the keys
No standing remote access
We don't keep a way into your machine. If remote help is ever arranged, it requires your approval and your presence. For everyday questions, Scout, our AI support agent, is available in chat.
Updates on your terms
ProductiveBot app updates are signed, and you install them when you choose from inside the app. We don't quietly change your agent's setup after it's configured.
Report a security issue
If something looks wrong, email support@productivebot.ai with "Security" in the subject line.
ProductiveBot vs. a DIY setup
Ready for an agent with its own space?
Pick the Mac that fits your business. Every ProductiveBot comes set up, with its safeguards installed and Scout ready to help.