Privacy policy

Who we are

Fullsight is run by Ernst Development LLC. It helps companies see and manage the AI platform accounts they pay for. This policy covers the information Fullsight handles and what we do with it. Questions go to ethanernsttt@gmail.com.

What we collect

• Your account: your name, email address and password (stored only as a one-way hash), and the organizations and roles you belong to. If you turn on two-factor sign-in, its secret and backup codes too, encrypted.
• From platforms your organization connects: what each platform's admin API reports, such as member names and email addresses, groups and who's in them, seats and plans, usage and cost, spend limits, audit logs, skills, plugins, connectors, MCP servers and service accounts. For API keys, that means each key's name, owner, dates and the masked version the platform shows, never the key itself. We don't ask for the contents of people's conversations.
• What your team adds: directory files, lists of seats, groups, skill packages, MCP server details, budgets, seat prices, settings, approval decisions and comments.
• Secrets: platform admin keys and a Slack webhook address, if you add them. They're encrypted with AES-256-GCM when saved and never shown again; the app only displays their last 4 characters.
• Technical information: IP addresses, used to limit repeated sign-in attempts; the IP address and browser details stored with each sign-in session; and the cookies that keep you signed in. We don't use analytics or advertising cookies or scripts.

How we use it

Only to run Fullsight for your organization: to show your data to the people in it, to send the email and Slack alerts your organization turns on, to keep its audit log, and to protect the service from abuse.

We don't sell your information or use it for advertising, and nothing is sent to an AI model unless your organization's owner turns on an AI feature (below). An organization's data is only shown to its own members, according to their roles: other people's seats, usage, cost and API keys are visible only to owners and admins, and everyone else sees company totals and their own.

AI features. Two optional features use Claude, made by Anthropic. Both are off until the owner turns them on in Settings › AI, and each runs only when someone in the organization asks:

• An AI review of a skill or an MCP server sends the skill's text files and the results of our automatic checks, or the server's configuration (with only the host of its address) and tool list.
• A monthly report summary sends that report's totals by platform and team. It never includes a person's name or email address, though the total for a very small team can reflect one person's activity.

Each request is recorded in the organization's audit log. Results are stored with the organization's other data until it's deleted.

Connecting your own AI tool. You can connect an AI tool to Fullsight over its MCP server. It then reads, as you, the shared skills, approved MCP servers, and skills and servers sent to you, that you ask it about, and that content goes to that tool's provider under your agreement with them. If you ask it to send something, it can look up colleagues' names and email addresses to prepare that, and you confirm the send in Fullsight. It can't see other people's seats, usage or cost. Nothing is connected unless you approve it, and you can disconnect it in Settings › AI tools.

Services that process it for us

• Vercel hosts the app.
• Neon hosts the database, in the United States.
• Resend sends email: confirmation and password reset links, invitations, alerts (including when something is sent to you, which you can turn off from any alert email), digests and reports.
• Inngest runs scheduled jobs such as syncs and weekly digests, and receives the identifiers and results of those jobs.
• Sentry receives reports of errors: what failed, where in the code and on which page. Email addresses and known key formats are removed from the error text first, and we don't send who was signed in or what was sent with the request.
• Slack receives alert messages, only if your organization adds a Slack webhook.
• Anthropic receives what an AI feature sends (above), only if your organization's owner turns that feature on, and returns the review or summary.
• The AI platforms you connect receive our read requests, made with the keys your organization provides.

Keeping and deleting it

We keep an organization's data while it exists, unless its owner sets how long usage, cost and audit history is kept in Settings › Security; older history is then deleted daily. An owner can delete the organization and everything in it at any time in Settings › Danger zone; that also removes its members and invitations, but not people's own accounts. To delete your own account, email ethanernsttt@gmail.com from that address and we'll do it. Copies in our database provider's backups expire on that provider's schedule.

Removing a platform connection, or deleting all credentials in Settings › Danger zone, deletes the stored key and stops further reads. To make sure a key can't be used again, also revoke it in that platform's admin console.

Security

• Keys and webhooks: each one gets its own random data key and is encrypted with AES-256-GCM. That data key is itself encrypted with a key kept outside the database, so a copy of the database alone doesn't open them.
• Between organizations: everything someone does in the app is limited to their organization, and the database enforces that limit again with row-level security. Background jobs such as syncs are limited to one organization by our code.
• Signing in: new accounts must confirm their email address, and the password is chosen from that emailed link, so nobody can take over an address before its owner signs up. Repeated failed attempts are limited, and anyone can add two-factor sign-in to their account with an authenticator app. Owners and admins can limit invitations to their email domains and set how long a sign-in lasts. That limit doesn't apply to a connected AI tool, whose access renews for up to 30 days unless it's disconnected or the person is removed.
• Inside your organization: changes are recorded in an audit log that owners and admins can read and export. Removing someone ends their access, and any AI tool they connected, checked again on every request.

No system is perfectly secure. If we learn of a breach affecting your data, we'll tell your organization's owners. To report a security problem, write to ethanernsttt@gmail.com.

Children

Fullsight is a tool for work and isn't meant for children.

Changes

If this policy changes, we'll update the date at the top, and we'll email organization owners about significant changes before they take effect.