Security
An agent gets exactly the access you grant it — and nothing else on your network.
Every agent works on its own computer under a security profile you choose — from an open setup computer to no internet, a read-only filesystem, and no credentials injected at all.
The Security Problem
Letting an AI act inside your systems shouldn’t mean handing it the keys to all of them.
An assistant that answers questions is easy to trust. One that clicks, types, and files inside your fleet software, your mailbox, and your billing system is a different question.
Real Risks
One login into everything
A single credential set that reaches every system, used by something nobody is watching.
Actions land in live systems
An agent that sends mail, updates a load, or writes a record is changing something your customers see.
Credentials in the conversation
A password pasted into a chat box ends up in a transcript, a log, and a model’s context.
No record of what happened
When something goes wrong in the yard system on Tuesday, nobody can reconstruct which step did it.
Regulated data in the way
Mailboxes, charts, and billing records carry PHI and PII. A general-purpose assistant has no scope drawn around them.
Honesty note: Isolation reduces risk, it doesn’t erase it — an agent you grant a system can still make a mistake inside that system. We say isolated, scoped and recorded, never safe.
Isolation, by default
Every agent works on a computer of its own.
Instead of handing an AI a shared login on a staff computer, AptiveAI gives each agent its own computer — a browser and a desktop it can use freely, holding only the access you granted, on a screen you can watch.
Its own computer
Every agent gets a computer with its own browser and desktop — not a login on somebody’s laptop.
Only the connectors you granted
An agent holds credentials for the systems you connected it to, and nothing else on your network.
Blast radius of one computer
A mistake is contained to that computer and those systems. Other agents and the rest of your network are out of reach.
You can watch the screen
Live view streams the agent’s browser and desktop while it works. Take over mid-task whenever you want to.
Everything is written down
Every action and every tool call lands in the run transcript, kept for the term of your contract.
Before
Every system, one set of credentials, and no record of who did what.
With AptiveAI
Each agent on its own computer, with its own credentials, on a screen you can watch.
Security Profiles
From an open setup computer to no internet at all — one setting.
* Hard ceilings on RAM (every profile) plus CPU cores and process count (Standard and up), enforced by the kernel — one runaway agent can't exhaust the host or starve the agents next to it.
- Network
- Full internet
- Filesystem
- Read-write
- Credentials
- Injected
- Hardening
- default
- Runaway-process protection*
- Generous4 GB RAM · CPU/PIDs uncapped
- Best for
- Setting an agent up
- Network
- Connected systems only
- Filesystem
- Read-write
- Credentials
- Injected
- Hardening
- cap-drop ALL, no-new-privileges
- Runaway-process protection*
- Tuned2 CPU · 4 GB · 256 procs
- Best for
- Everyday runs
- Network
- Your allowlist only
- Filesystem
- Read-only + carve-outs
- Credentials
- Injected
- Hardening
- cap-drop ALL, no-new-privileges
- Runaway-process protection*
- Lean1 CPU · 2 GB · 128 procs
- Best for
- Customer records, billing
- Network
- No internet
- Filesystem
- Read-only + carve-outs
- Credentials
- Not injected
- Hardening
- cap-drop ALL, no-new-privileges
- Runaway-process protection*
- Lean1 CPU · 2 GB · 128 procs
- Best for
- On-premises and regulated work
* Hard ceilings on RAM (every profile) plus CPU cores and process count (Standard and up), enforced by the kernel — one runaway agent can't exhaust the host or starve the agents next to it.
Network Egress Control
Allowed ✓
outlook.office365.comgraph.microsoft.comfleet.yourcompany.comyard.yourcompany.com+ the systems you connectedfilter
Blocked ✕
file-drop.examplepaste-my-data.iorandom-exfil.netunknown-api.xyzeverything not on the listYour agent can reach your fleet system. It cannot reach the open internet.
Hostnames that are not on the list return NXDOMAIN — customer records, documents, and credentials have nowhere to go.
Trust & compliance
How we run the service itself.
The controls above protect each agent run. These are the operational practices that protect the console, your account, and the data we hold on your behalf.
Encryption in transit and at rest
Traffic to AptiveAI is served over TLS. The account, run, and telemetry data we store are encrypted at rest.
Credentials live in the credential store
Connector credentials are held in the credential store and handed to the computer at run time. They are never written into the transcript and never shown in the live view.
Least-privilege access
Scoped access controls and continuous monitoring govern who and what can reach our production systems. An agent reaches only the systems you connected it to.
Incident response
If a breach affects you, we notify you without undue delay and notify the relevant supervisory authority within 72 hours where the law requires it.
Vetted sub-processors
Clerk (identity) and Microsoft Azure (US hosting, storage, and transactional email) — each under contract that restricts how it may use your data.
Honest compliance posture
We hold no SOC 2, ISO, or HIPAA attestation and sign no BAA today. Agents act inside live systems, so a workload touching PHI or other regulated data is a compliance conversation before it is a contract — and it can run on your own infrastructure instead of ours.
Found a vulnerability?
We welcome responsible disclosure. Email security@aptiveai.io. For a question about the data we hold, write to privacy@aptiveai.io; for anything else, support@aptiveai.io.
Every computer runs under a security profile — you choose the level.
Network egress control keeps an agent to the systems you connected.
Read-only filesystem on Strict and Offline; privilege hardening on Standard and up.
Offline profile has no internet and injects no credentials — for on-premises and regulated work.
With AptiveAI: An agent gets exactly the access you grant it — and nothing else on your network.