Privacy Policy for toolzilla.com

Toolzilla is a hosted gateway for the Model Context Protocol (MCP). It gives an organization one endpoint for its AI agents, with scoped access tokens, stored credentials for the services it connects, and a record of tool calls.

Defcon Systems, LLC (“we”, “us”) operates Toolzilla and the toolzilla.com website. This policy explains what information we handle, why, who receives it, how long we keep it, and what you can do about it.

This section covers what we receive from Google when you use Sign in with Google.

• What we receive: your name, email address, profile photo, and Google account ID, through the email and profile scopes. Signing in gives us no access to your Gmail, Drive, Calendar, or other Google content.
• How we use it: to create your account, sign you in, show your name and photo in the product, send you account email, and keep the service secure.
• How we store it: with our database and authentication provider, with your photo copied to our object storage, until your account is deleted.
• Who we share it with: only the service providers in section 8, as needed to run the service. Our staff can see it as section 7 describes, and your display name can accompany request content sent to an AI provider, as section 6 describes. We do not sell it, use it for advertising, or use it to train AI models.
• How to remove access: revoke access in your Google Account permissions, unlink Google in your account’s sign-in settings, or ask us to delete your data as section 10 describes.

Toolzilla’s use and transfer of information received from Google APIs adheres to the Google API Services User Data Policy, including the Limited Use requirements where they apply to the scopes we request.

Google connectors work differently. Connectors such as Google Drive, Google Calendar, and BigQuery do not use Sign in with Google. An organization gives us a key for a Google Cloud service account it controls. We store the key encrypted and use it only for the tool calls that organization’s agents make. The content those calls return goes to the agent that asked for it, and we keep only the tool call record section 5 describes.

• To provide, operate, and support the service, including signing you in, routing tool calls, and applying plan limits.
• To handle subscriptions and billing.
• To send service email, such as invitations, confirmations, and notices about your account.
• To understand how the service is used, through internal analytics, and to improve it.
• To detect and prevent fraud, abuse, and security problems.
• To comply with legal obligations.

We do not sell your personal information or use it for advertising.

Where the GDPR or UK GDPR applies, our legal bases are performing our contract with you, our legitimate interests in running and securing the service, your consent where we ask for it, and our legal obligations.

When an agent calls a tool, the request passes through our servers to the connected service, and the answer passes back to the agent.

• Normal calls: we keep a record of the tool, the connection and token used, the status, time, and duration, diagnostic details with sensitive content removed, and a description of the arguments: their type, length, a hash, and their field names. We do not keep the argument values or the response.
• Long-running calls: we hold the arguments encrypted until the job finishes, then clear them. We store the full result for 30 days after the job finishes, so your agent can retrieve it, then delete it. The service running the job can hold its own copy on its own schedule.
• Events: event content from connected services is stored encrypted and removed on the schedule in section 9.

You choose your agents, the AI providers behind them, and the services you connect. What they do with the data they receive is governed by your agreements with them, not this policy.

When you file a request through the Requests feature, an AI agent helps our staff handle it. It runs on Anthropic’s Claude, and approved delivery work can use other AI model providers.

• Without any staff step, the agent sees your organization name and the names of your connections and credentials.
• After a staff member admits them, it can also read request briefs, messages, feedback, repository snapshots, and their authors’ names, plus data from any connection shared for that request.
• A staff member approves anything sent to you or that moves a request forward. The agent can write internal notes and labels on its own.

We do not use your information to train AI models. These providers handle what they receive under their own terms.

Our staff use access to customer information to run, secure, and support the service.

• Staff administrators can read requests across organizations and download the repository snapshots attached to them.
• Staff receive Slack notifications with the email address and company from access requests, plus request titles and message excerpts of up to 600 characters.
• An internal analytics dashboard shows member email addresses, roles, last sign-in times, organization names, and usage counts.
• Engineers with production database access can read stored job results and request messages. They use that access only to operate or secure the service, or to resolve a problem you report.

We share information with these providers to run the service:

• Supabase: database and authentication.
• Cloudflare: application hosting, file storage, background jobs, network delivery, and security checks. Its Turnstile check on some forms receives your IP address.
• Vercel: website hosting.
• Stripe: payments and subscriptions.
• Postmark: service email.
• Slack: internal staff notifications.
• Google Workspace: our support email, including data deletion requests.
• Anthropic and other AI model providers: the AI features in section 6.

We also pass information to the services you connect, as you direct. We may disclose information when the law requires it, to protect rights and safety, or as part of a merger or sale of the business, with notice to you.

• For the life of your account or organization: account information, requests and their messages, and sent invitations. These have no automatic expiry.
• Job results: 30 days after the job finishes.
• Tool call records: 13 to 14 months after the call.
• Until you ask us to remove them: waitlist entries and access requests.
• Connection credentials: erased when you remove the connection. We keep a record that the connection existed, so tool call records still make sense.
• Repository snapshots: 7 days after the request closes or 30 days after creation, whichever comes first, or sooner if you disconnect the repository access.
• Events: event content and delivery records for 7 days, and ended subscriptions for 30 days.
• Sign-in history: 400 days.
• Invitations awaiting signup: 30 days.
• Tokens from OAuth clients: access tokens for 1 hour and refresh grants for 30 days.

Copies in our providers’ backups expire on their schedules.

To delete your account and personal information, email support@toolzilla.com from the address on your account. We will confirm it is you, then delete or de-identify your information within 30 days. We keep what we must for billing, tax, legal, and security reasons. Records that belong to your organization can remain with it, no longer tied to you. If you are an organization’s only owner, we will ask what should happen to it.

You can also remove connections, revoke tokens, unlink sign-in providers, and leave an organization yourself in the product.

Depending on where you live, you may have the right to access, correct, delete, or receive a copy of your personal information, and to object to or restrict how we use it. To use these rights, email support@toolzilla.com. You may also complain to your local data protection authority.

We use only the cookies the service needs. We do not use advertising or analytics cookies.

• Session: keeps you signed in to the dashboard.
• Short-lived flow cookies: carry an invitation, a security re-check, or a connection you are setting up, and last from minutes to about an hour.
• Preferences: remember interface choices, such as the sidebar, for up to a year.

We do not track you across other websites, so we do not change what we do in response to browser Do Not Track signals.

In the past 12 months we have collected identifiers (such as name and email address), commercial information (such as subscription status), internet activity (such as tool call records), and what you send us, for the purposes in section 4, from you and the sign-in providers you choose.

We do not sell personal information or share it for cross-context behavioral advertising, and we do not use sensitive personal information beyond running the service. California residents can ask to know, correct, or delete their personal information, directly or through an authorized agent, by emailing support@toolzilla.com. We will not treat you differently for asking.

Questions and requests: support@toolzilla.com

Defcon Systems, LLC. 113 West G Street, #549. San Diego, CA. 92101. USA