Privacy Policy

VAIL Agent Assurance Engine — Privacy Policy

Version: 2026-11-11

This Privacy Policy explains how VAIL, Inc., a Delaware corporation with offices located at 1027 Hayes Street, San Francisco, CA, United States (“VAIL,” “we,” “us,” or “our”) collects, uses, retains, and protects information in connection with the VAIL Agent Assurance Engine web application, MCP server, connectors, and related developer tools (collectively, the “Services”). The Services are intended for business and developer use and are governed by our Terms of Service.

1. Information We Collect

1.1 Account and Organization Information. When you create an account, accept a team invitation, sign in, or connect a coding agent or MCP client, we collect business contact and account details such as your name, work email address, company or organization name, organization role, Terms of Service acceptance record, sign-in timestamps, and IP address used for rate-limiting and abuse prevention.

1.2 Authentication Credentials and Connected Clients. When you sign in through the web application, create personal access tokens, or authorize an OAuth 2.1 MCP client (such as Claude, Claude Code, ChatGPT, Codex, or OpenCode), we store cryptographic hashes of access tokens, refresh tokens, and authorization codes together with client identifiers, token names, scopes, creation times, expiration times, and last-used timestamps. If an agent endpoint credential is configured for live probing, secret material is stored encrypted in AWS Secrets Manager rather than in plaintext application records.

1.3 Agent Materials and Probe Data. When you upload source files, connect a GitHub repository through the VAIL GitHub App, scan an agent, generate probes, or run probes against a connected agent, we process the files, prompts, tool schemas, generated behaviors and probes, and agent responses (“Customer Data”) needed to provide the Services. As required by our Terms of Service, you should not upload Personal Information of end consumers, protected health information, payment cardholder data, or plaintext production secrets.

1.4 Service Usage and Technical Logs. We record operational logs, job status, token-spend rollups, and error reference identifiers needed to operate, secure, and troubleshoot the Services. We do not use third-party advertising trackers on the Services.

2. How We Use Information

2.1 Providing and Securing the Services. We use the information we collect to authenticate users and authorized MCP clients, isolate each organization’s workspace, analyze agent definitions, generate and grade behavioral probes, send account and team emails (such as magic-link sign-ins and invitations), enforce rate limits and spending controls, and provide customer support.

2.2 Model Providers and Subprocessors. When you run a scan, behavior classification, or probe generation, relevant Agent Materials and excerpts are transmitted to the configured model provider (such as Amazon Bedrock or OpenRouter) solely to perform that analysis. As set forth in our Terms of Service, we only use model providers committed not to train their models on Customer Data and to maintain industry-standard security safeguards. We also use Amazon Web Services (AWS) in the United States to host application infrastructure, databases, encrypted secrets, and transactional email.

2.3 No Sale of Information or Model Training. We do not sell personal information, share it for cross-context behavioral advertising, or use your private Agent Materials or Customer Data to train foundation models.

3. Data Isolation, Retention, and Deletion

3.1 Organization Isolation. Each organization’s agents, uploaded source files, generated artifacts, and probe runs are scoped to that organization. Authorized OAuth tokens and personal access tokens act only as the user who authorized them within their organization and stop working immediately if revoked or if that user leaves the organization.

3.2 Retention and Deletion. Organization owners can delete any agent at any time through the Services, which removes its collected source files, probes, schedules, findings, and stored endpoint credentials, subject to probe run history retained for up to 180 days (or the deployment’s configured retention window) and routine encrypted backups deleted in the ordinary course (typically within 30 days). You may revoke personal access tokens and OAuth client connections at any time under Access tokens in the web application. To request deletion of your account or organization records, contact us at privacy@projectvail.com or contact@projectvail.com.

4. Security

We encrypt data in transit over TLS and at rest using industry-standard encryption, store secret tokens as cryptographic hashes or in dedicated secrets management, enforce per-organization access checks on every web and MCP request, and restrict staff access as described in Exhibit A of our Terms of Service.

5. Your Choices and Contact Information

You may review your team membership, revoke connected MCP applications and access tokens, or delete agents directly in the Services at https://agents.projectvail.com. For privacy questions, data deletion requests, or exercises of applicable statutory privacy rights, contact VAIL, Inc. at privacy@projectvail.com or contact@projectvail.com, or by mail at 1027 Hayes Street, San Francisco, CA, United States.