Privacy Policy
This page describes what data SentinelX collects, why, and how it's stored. SentinelX is a hosted MCP service that lets you operate your Linux, macOS and Windows servers from an AI client (such as ChatGPT, Claude, Codex or Cursor) through a small agent installed on each server.
What we collect
Account data
When you sign in with Google or GitHub, we receive from your identity provider:
- Your email address
- Your name (as provided by your identity provider)
- A stable user ID issued by your identity provider (the OpenID Connect
subclaim)
We use these to identify your SentinelX account and bind your enrolled hosts to it. We do not receive your password — your identity provider (Google or GitHub) handles authentication.
When you sign in, we also record the approximate country you connect from, derived from your IP address. We store the country, not the IP address.
Host data
For every server you enroll with SentinelX, we store on our hub:
- A unique host ID generated at install time
- Your server's hostname (as reported by the agent at handshake)
- Any optional label you've set
- The agent version and protocol version
- Metadata about the active connection (session ID, connect time)
- The approximate country the agent connects from, derived from its IP address (we store the country, not the IP address)
We do not store the contents of your server's filesystem, the output of commands you run, or any application data on your host. The agent runs commands locally and returns results to your LLM session — they pass through our hub but are not persisted.
Operational logs
For security auditing, service health, debugging and abuse prevention, our hub records an audit entry for each tool call, including:
- Your user ID (not your email)
- The target host ID
- The operation name (
exec,edit, etc.) and the arguments your AI client sent with it, for example a command line, a file path, the text written by an edit, or the contents of a script - A timestamp, and whether the call succeeded (with the error message if it didn't)
- Timing, the size of the response and a cryptographic hash of it (not the response itself)
- The name and version of your AI client, and request and session identifiers
The output of commands and the contents of files you read through SentinelX pass through our hub but are not stored. Don't put secrets in command lines or scripts: they are recorded like any other argument.
If you share a host with another SentinelX account, its audit entries record which account performed each operation.
What we do NOT collect
- Your password — your identity provider authenticates you, we never see it
- Output of commands you run, or contents of files you read, through SentinelX
- SSH keys, credentials, or other secrets stored on your hosts
- Browsing or telemetry data unrelated to using SentinelX
How long we keep data
- Account data: for as long as your account exists. Delete your account to remove it.
- Host registrations: until you uninstall the agent or revoke the host.
- Operational logs: 30 days, then automatically rotated out.
Who we share data with
We do not sell your data. We share account data only with infrastructure providers required to operate the service (compute, DNS, certificate authorities), and only as needed for them to deliver their service. Traffic to our hub passes through Cloudflare. If you subscribe to a paid plan, payments are processed by Polar (with Stripe as its payment provider), which receives the billing details needed to charge you.
Where data is stored
The SentinelX hub and its databases run on DigitalOcean servers in New York, United States. If we add or change regions, we'll update this page.
Your controls
- Disconnect a host: uninstall the agent (instructions for Linux, macOS and Windows in the installer repository) or remove its identity file. The host stops appearing in your account.
- Delete a label: ask your LLM to call
sentinel_remove_host_label. - Delete your account and all associated data: email privacy@sentinelx.app. We process the request within 30 days.
Security
All connections between your agent, our hub, and your LLM are TLS-encrypted. Identity tokens are short-lived JWTs. The agent never accepts inbound connections — it only opens an outbound WebSocket to our hub. Each agent's authority is scoped to its own host.
Changes to this policy
We'll update the "last updated" date when we make material changes. For significant changes, we'll email you via the address provided by your identity provider.
Contact
Questions: privacy@sentinelx.app