Privacy policy

Last updated: August 18, 2026

1. Who We Are

PerilScore is operated by Peril Score, LLC ("we," "us," or "our"). We provide property-risk intelligence at app.perilscore.com, through APIs, and through user-authorized integrations such as the ChatGPT App and Model Context Protocol (MCP) connectors for Claude and other compatible clients.

2. Information We Process

Depending on the feature you use, we may process:

  • Property inputs and results, including addresses, coordinates, property characteristics, construction and COPE data, hazard scores, replacement-cost estimates, permit signals, imagery-derived observations, reports, and information you include in a statement of values.
  • Account and organization data, including name, email address, organization membership, plan and usage information, saved scores, and report history.
  • Integration and authorization data, including MCP tool names and arguments, OAuth client and grant metadata, authorized scopes, access or refresh token records, API-key identifiers, and the PerilScore account connected to an integration.
  • Billing and communications data, including billing-customer identifiers, transaction status, and records of transactional emails or messages. Payment card details are handled by our payment processor rather than stored by PerilScore.
  • Technical and security data, including IP address, browser or client type, timestamps, request and error logs, referral source, and abuse-prevention signals.

A property address can relate to a person even when it is submitted for commercial property analysis. Do not submit personal or confidential information that is unnecessary for the requested analysis.

3. How We Use Information

We use information to provide and secure the Service, authenticate users and integrations, perform property enrichment and scoring, generate reports, administer usage and billing, deliver transactional communications, troubleshoot errors, prevent abuse, and improve model and service performance. PerilScore provides non-binding decision-support evidence; it does not make final insurance coverage decisions.

4. ChatGPT, Claude, and MCP Integrations

When you invoke PerilScore from ChatGPT, Claude, or another MCP-compatible client, that client sends us the requested tool name, the fields needed to run it, protocol and security metadata, and, for protected tools, information identifying the authorized PerilScore account and scopes. We return the requested result to that client. We do not request or retrieve unrelated conversation history, memories, files, or account data from the client, although text you intentionally place in tool arguments is processed as part of the request.

Free score-only tools may not require a PerilScore account. Premium tools can require OAuth authorization or another supported credential and may process property enrichment, construction and COPE data, indicative replacement-cost estimates, permit records, imagery observations, and AI-generated summaries. You can revoke an OAuth authorization through the available PerilScore account or authorization controls; revocation prevents future authorized use but does not by itself erase records already created.

Your use of a third-party client is also governed by that provider's privacy terms. The client decides how it stores conversations and tool results after PerilScore returns them.

5. Service Providers and Data Sharing

We do not sell personal information. We may disclose information to legal authorities when required by law or necessary to protect rights and safety, and to provider categories needed to operate a requested feature:

  • Hosting and database infrastructure, currently including DigitalOcean for the production application and database.
  • Geocoding, maps, routes, places, and imagery services used to resolve a property and provide location or building context.
  • Property-data and building-permit providers used for premium enrichment when data is available.
  • AI model providers used for identified imagery observations or narrative summaries. AI output is not the sole basis for the deterministic peril scores.
  • Email, messaging, payment, monitoring, and security providers used for account operations, billing, delivery, reliability, and abuse prevention.
  • User-selected integration providers, such as OpenAI or Anthropic, when a user asks that client to call PerilScore and receive the result.

Operator confirmation required before Claude directory launch: PerilScore must verify the live provider inventory and confirm that each applicable property, permit, imagery, and AI provider agreement permits the intended MCP processing and delivery of results to user-selected AI clients. This policy does not claim redistribution rights that have not been contractually confirmed.

6. Retention and Deletion

We retain account, authorization, score, report, billing, security, and technical records only for as long as reasonably needed for the purposes described above, including providing report history, maintaining security and audit records, meeting legal obligations, resolving disputes, and enforcing agreements. OAuth access and refresh tokens follow their configured expiration and revocation behavior. Deletion from active systems may not immediately remove data from backups or records we must retain for legal, fraud-prevention, billing, or security purposes.

Retention schedule under operator confirmation: exact production periods for anonymous address requests, account score history, application and security logs, OAuth records, provider caches, and backups have not yet been published. Those periods must be verified against live infrastructure and provider settings before Claude directory launch; we will update this policy rather than inventing a fixed period.

You may request deletion of your account and associated personal data by emailing [email protected]. We will verify the request and explain any information that cannot yet be deleted and why. Revoking an integration and requesting deletion are separate actions.

7. Security

We use safeguards designed to protect information, including HTTPS in transit, access controls, credential-scoped authorization, and passwordless account authentication. No security measure is perfect. Please report a suspected vulnerability to [email protected] with the subject "Security Report" and review our responsible-disclosure guidance.

8. Your Choices and Rights

Subject to applicable law, you may request access to, correction of, or deletion of your personal information. You may also revoke integration access, stop using anonymous tools, or contact us with a privacy objection or question. We may need to verify your identity or authority over an organization before acting on a request.

9. Children and Changes

The Service is intended for business and professional use and is not directed to children. We may update this policy as our services and provider relationships change. The date above identifies the current version.

10. Contact

For privacy requests, deletion requests, or security reports:

Peril Score, LLC
Email: [email protected]