Infrastructure Security
Meeting Assistant is pending Zoom Marketplace approval and coming soon; it is not generally available yet. The technical descriptions below describe controls for use if approved and enabled.
The managed OpsPing cloud runs as one active production deployment in AWS us-east-2 (Ohio), backed by PostgreSQL on the application host. A stopped, pre-provisioned recovery environment in us-east-1 (N. Virginia) can be promoted by an operator; there is no automatic failover. Email is delivered via AWS SES. The hosted Meeting Assistant runs single-use Fargate bot tasks and stores transcripts in a private S3 bucket in AWS us-east-1; summary generation, summaries, and meeting metadata run in the core PostgreSQL deployment in us-east-2, with inference on Cloudflare Workers AI's global network. Public marketing, blog, and status sites are hosted on Cloudflare Pages; billing runs through Stripe. Private single-tenant deployments run the same application on infrastructure you or we manage.
Data Encryption
- In transit: All communication uses TLS 1.2+, terminated by Caddy in front of the application. HTTP requests are redirected to HTTPS.
- At rest: Production and per-customer EC2 root volumes use AWS KMS-backed EBS encryption. Encrypted recovery copies of database backups and archive logs, attachments, and configuration are kept in us-east-1. Sensitive fields — channel values (on-call phone numbers and email addresses), MFA secrets, and Slack integration tokens — are additionally encrypted at the application layer with AES-256-GCM. Meeting Assistant transcripts are stored in a private S3 bucket in us-east-1 with SSE-KMS; Zoom OAuth tokens are AES-256-GCM encrypted. Card and bank data is entered in Stripe-hosted Checkout or the Customer Portal and never enters OpsPing systems.
- API keys: Stored as bcrypt hashes in the database. No plaintext API keys exist in our database. After initial creation display, keys cannot be retrieved — only rotated.
Network Security
- The public API and admin surface terminate at Caddy on the application host. It terminates TLS, redirects HTTP to HTTPS, sets security headers, and proxies to the service listening on localhost. The marketing, blog, and status sites are served from Cloudflare Pages, and the Meeting Assistant inference endpoint is a Cloudflare Worker at
ai.opsping.com. - The application talks to SES and S3 over TLS to AWS service endpoints, to the PostgreSQL database locally on the host, and to push/voice providers (APNs, FCM, Twilio) over HTTPS. SMS is not offered in new channel setup; legacy SMS channels remain deliverable only if already enabled. No third party has a network path into the database.
- Meeting Assistant bot tasks are single-use and accept no inbound connections (their security group denies all ingress). Each task receives only its per-session HMAC-signed token, the short-lived Meeting SDK JWT and On-Behalf-Of join token, the meeting passcode, and a shared, static, route-scoped bearer token for the transcription endpoint — no long-lived Zoom OAuth tokens and no general cloud credentials. Audio is sent over TLS to the
/transcriberoute of theopsping-meeting-summaryWorker atai.opsping.com, which calls Cloudflare Workers AI, and held only in ephemeral task storage until the pipeline finishes. - SSH access is key-only — password authentication is disabled. Application-level rate limiting guards authentication and alert-ingestion endpoints.
Access Control
- IAM: The EC2 instance runs under an IAM instance role scoped to SES sending and S3 attachment storage — least privilege, no standing console credentials on the host, no database credentials in IAM.
- API auth: API keys (OpsPingKey / GenieKey-compatible) for programmatic access, and signed JWTs for mobile/web sessions. Both are validated on every request.
- Infrastructure access: Server changes follow a documented deploy runbook. Per-customer (single-tenant) infrastructure is provisioned with OpenTofu (
tofu/aws-silo). - Employee access: Production host access is limited to infrastructure administrators using individual SSH keys and documented break-glass procedures. Support has no application-level access to customer data unless invited into the customer's organization.
Application Security
Authentication
- Passwords are hashed with bcrypt. TOTP-based MFA with backup codes is available, and admins can reset a user's enrolled factor.
- API keys are random tokens with per-key scoping (read, write, delete, config) and optional binding to a specific team. Keys are stored only as hashes.
- JWT access tokens expire after 15 minutes by default; refresh tokens last 30 days with rotation. Password resets, blocks, and role changes invalidate all outstanding tokens immediately.
- Account recovery requires verified email. Authentication and registration endpoints are rate-limited.
Input Validation & Output Encoding
- All API inputs are validated server-side by the Go backend's typed request schemas. Unknown fields are rejected (HTTP 400), not silently stored.
- Dynamic content rendered in the admin dashboard is escaped via React's default JSX escaping.
- API responses set
Content-Type: application/json— no content sniffing possible. - CSP headers configured on the admin dashboard and marketing site.
Dependency Management
- Dependencies are pinned: exact versions in
package-lock.json(clients and tooling) andgo.sum(backend). - Automated Dependabot alerts for known CVEs; our target is to apply critical patches within 48 hours of availability.
- Runtime: the API is a compiled Go static binary (go 1.27 per
go.mod); client build tooling runs on Node.js LTS.
Zero Standing Access
OpsPing support has no application-level access to your data. Support can see your alerts only when you invite the support account into your organization, and everything it does is recorded in your own audit log. Infrastructure administrators retain technical access to production systems for operations and recovery; that access is limited to documented break-glass procedures.
Tenant Isolation
- Every organization's data is scoped to a tenant id. Cross-tenant access is denied by default — there is no code path that lets staff read another customer's alerts, teams, or users.
- The same isolation machinery guards every request; staff accounts get no bypass.
Support Access by Invitation
- When you need hands-on help, you invite support@opsping.com into your organization through the normal invite flow, with the least role needed for the task.
- You can revoke access at any time by deleting the user or letting the invite lapse.
- Every action support takes while invited lands in your audit log, attributed to the support account — nothing happens off the record.
Break-Glass Accountability
- Manual database intervention is a documented, exceptional runbook — never routine support.
- A point-in-time backup is taken before any manual intervention.
- Wherever possible the work itself is done through the normal audited UI (via a temporary user inside your tenant) so your audit log records it; direct item edits are a last resort and are documented with exact before/after values.
- Every intervention ends with an incident record: what was done, why, by whom, and when.
Compliance & Certifications
SOC 2
OpsPing is not pursuing SOC 2 certification — no independent audit has been conducted and none is planned. Our practices are mapped to the SOC 2 Trust Services Criteria, in depth, on our SOC 2 page.
GDPR
OpsPing acts as a processor for customer data and offers a Data Processing Agreement (DPA) with the 2021 EU Standard Contractual Clauses (Module Two) for customers who need one. The core application, primary database, backups, the Go summary pipeline, summaries, and meeting metadata are stored/run in AWS us-east-2 (Ohio, USA), with encrypted recovery copies in AWS us-east-1; Meeting Assistant bot tasks and the transcript S3 bucket run in AWS us-east-1 (N. Virginia, USA); summarization inference runs on Cloudflare Workers AI. Cloudflare's global network handles DNS, edge/Pages hosting, Email Routing and the Email Worker for inbound alert email, and Workers AI inference (including public-site request metadata; the meeting Worker has no regional or jurisdictional pin configured); Stripe, LLC processes billing in the United States with global onward transfers under its DPA/SCCs; human mailbox correspondence is hosted by Proton AG in Switzerland, which is covered by the European Commission's adequacy decision for Switzerland, so no further transfer safeguard applies to that hosting. Transfers to OpsPing are covered by the DPA's SCCs; Cloudflare's, Stripe's, Google's, and Apple's own data processing terms and SCCs govern their onward processing (push delivery uses United States provider accounts and their global edge networks); any onward transfer by Proton outside Switzerland, the EU, or another adequacy country relies on Switzerland-, EU-, or UK-approved standard contractual clauses or another lawful transfer mechanism under Section 11 of Proton's Data Processing Agreement. See our Privacy Policy for details on data handling, retention, and deletion.
Data Retention
- Closed alerts, the audit trail, and other operational data: retained for 90 days, then automatically purged.
- Open alerts: retained until closed.
- Account and channel data: retained while the account is active; deletion requests are honored within 30 days of a verified request.
- Meeting Assistant audio: held only in the bot task's ephemeral storage and deleted when the transcription pipeline finishes. Transcripts in S3 us-east-1 and summaries/metadata in PostgreSQL us-east-2 are deleted by the daily retention job when the organization's window (1–365 days; default 30) elapses. The job deletes each expired transcript object from S3 before deleting its row; when that object deletion fails, the row is kept and retried on the next daily sweep. The transcript bucket's separate S3 lifecycle rule expires objects after its configured period (fixed 400-day fail-safe: 365-day maximum tenant retention + 35-day margin) as a backstop. Deauthorization, console disconnect, and organization deletion delete the database records and attempt the stored transcript-object deletions; object-deletion failures are logged (with the object key when available) and the bucket lifecycle backstop removes any object left behind. Glossary entries and summary instructions are organization configuration (which may include admin-approved short excerpts from corrections) retained until deleted or the organization is purged, outside meeting-content retention. Slack copies follow the workspace's retention; internal incident-timeline copies follow the incident record's and organization's lifecycle.
- Billing records: retained while the account is active and as needed for tax/accounting; Stripe retains its own invoice and payment records.
- Inbound alert email follows alert retention; support/legal/security/sales/general-inquiry/migration/SMS-compliance/careers correspondence is retained for up to three years only where needed to respond, comply with law, or establish/exercise/defend legal claims. On termination of a customer agreement, ordinary Customer Personal Data in mailbox/support correspondence is deleted or returned within 90 days; only controller-side business/legal records lawfully required or needed for claims may be retained for that period, access-restricted, then deleted. Applicant/candidate recruiting records follow the same retention and deletion-request process.
- Backups: see the Privacy Policy.
Vulnerability Disclosure
If you discover a security vulnerability in OpsPing, please report it to security@opsping.com. We investigate all reports and respond within 48 hours. We don't run a public bug bounty yet, but we credit researchers in our engineering blog (with permission).
Do not attempt to access, modify, or delete other users' data when testing. Use your own account and test alerts only.
Questions?
Contact security@opsping.com for security questions, DPA requests, or penetration testing coordination. We're happy to provide additional documentation for your security review.
Enterprise Security Questionnaire
Evaluating OpsPing for your organization? This section answers the questions we most commonly receive in vendor security assessments. OpsPing is generally available. Some controls described here are still maturing; we disclose limitations plainly so your team can make an informed risk decision.
Product Overview
OpsPing is a SaaS on-call paging and alerting product for DevOps and operations teams, consisting of a React Native (Expo) mobile application and an AWS-hosted backend. Alerts are delivered via direct push notifications, optional voice calls (Twilio), and email (Amazon SES). SMS is not offered in new channel setup; legacy SMS channels remain deliverable only if already enabled.
Data Handling
| Question | Answer |
|---|---|
| What data do you process? | Account information (name, email, credentials), channel PII (on-call phone numbers and email addresses), mobile push tokens, alert logs and operational data (schedules, escalations, acknowledgements), and IP addresses / technical logs. Where the Meeting Assistant is enabled: meeting audio (ephemeral), participant display names and speaker labels, audio-derived transcripts (sent to Workers AI in chunks for summarization), prompts/instructions and organization summary instructions, partial and draft summary context, AI summaries and saved corrections, meeting metadata (including the meeting ID prepended to the transcript sent to Workers AI), linked incident record fields used for the terminology-correction pass (title, description, responders, impacted services, timeline, postmortem action items; assistant-posted timeline entries excluded; capped at 20,000 characters), organization glossary configuration (canonical term, definition, aliases; canonical terms used as transcription hotwords; terms/definitions/aliases used in summary-correction vocabulary; may include admin-approved short excerpts from corrections) and summary instructions, the authorizing Zoom user's user ID (we request only user:read:token and do not read Zoom profile, name, or email), bot-task CloudWatch logs (participant names/IDs, meeting timings), and encrypted Zoom OAuth tokens. Inbound alert email (sender/recipient, subject, body, Message-ID) where email-to-alert is enabled; support, legal, security, sales, general-inquiry, migration, SMS-compliance, and careers correspondence in Proton-hosted mailboxes, including applicant name, email, note, CV/application materials, and metadata. Billing: billing contact, Stripe customer/subscription identifiers, plan/seat state, and invoice/payment metadata; card and bank data stays in Stripe-hosted surfaces. |
| Where is data stored? | Core application, primary database, backups, the Go summary pipeline, summaries, and meeting metadata: Amazon Web Services us-east-2 (Ohio, USA). Meeting Assistant bot tasks and transcript S3 bucket: AWS us-east-1 (N. Virginia, USA), with SSE-KMS. Cloudflare hosts the public sites and runs Workers AI inference on its global network; the meeting Worker has no regional or jurisdictional pin configured, so inference may run outside the United States. Stripe, LLC processes billing in the United States with global onward transfers under its DPA/SCCs. Inbound alert email becomes alert content in the application database; Cloudflare Email Routing/Email Worker process it transiently. Human mailbox correspondence is hosted by Proton AG (Switzerland). |
| How long is data retained? | Closed alerts, the audit trail, and other operational data: 90 days, then automatically and permanently purged. Open alerts are retained until closed. Account and channel data are retained while the account is active. Meeting Assistant audio is held only in ephemeral task storage and deleted when the pipeline finishes; transcripts in S3 us-east-1 and summaries/metadata in PostgreSQL us-east-2 are deleted by the daily retention job when the organization's window (1–365 days; default 30) elapses. The job deletes each expired transcript object from S3 before deleting its row; when that object deletion fails, the row is kept and retried on the next daily sweep. The transcript bucket's separate S3 lifecycle rule expires objects after its configured period (fixed 400-day fail-safe: 365-day maximum tenant retention + 35-day margin) as a backstop. Deauthorization, console disconnect, and organization deletion delete the database records and attempt the stored transcript-object deletions; object-deletion failures are logged (with the object key when available) and the bucket lifecycle backstop removes any object left behind. Glossary configuration (which may include admin-approved short excerpts from corrections) persists until deleted or the organization is purged, outside meeting-content retention. Slack copies follow the workspace's retention; internal incident-timeline copies follow the incident record's and organization's lifecycle. Billing records are retained while the account is active and as needed for tax/accounting. Deletion requests are honored within 30 days of a verified request. Inbound alert email follows alert retention; support/legal/security/sales/general-inquiry/migration/SMS-compliance/careers correspondence is retained for up to three years only where needed to respond, comply with law, or establish/exercise/defend legal claims. On termination of a customer agreement, ordinary Customer Personal Data in mailbox/support correspondence is deleted or returned within 90 days; only controller-side business/legal records lawfully required or needed for claims may be retained for that period, access-restricted, then deleted. Applicant/candidate recruiting records follow the same retention and deletion-request process. |
| What about backups and recovery copies? | Encrypted backup and recovery copies may contain pseudonymous or customer identifiers; live erasure cannot rewrite already-created versioned backups. Those copies are access-restricted to recovery operations. Backup/recovery-copy expiry at or before 365 days and the separate 400-day tenant-lifecycle security-record retention are deployment-gated commitments, not verified current behavior: the canonical offsite lifecycle stack is not yet applied and verified live, and no 365/400-day maximum is asserted until release verification proves the live lifecycle rules. Receipts and backups are pseudonymous, not anonymous, and may remain linkable while the HMAC key or backups exist; access is key-restricted. |
| How are tenant deletion/restoration security receipts retained? | Append-only signed records contain only format version, delete/restore action, pseudonymous tenant identifier, timestamp, and HMAC; no names, email, Slack/Zoom identifiers, meeting content, or user identifiers. Purpose: keep deletion effective across restore and security accountability. Source AWS S3 opsping-offsite-970817257986 us-east-2, replicated to KMS-encrypted opsping-recovery-us-east-1-970817257986 us-east-1; no Object Lock. Planned retention is 400 days (399 current + 1 noncurrent) and planned backups/PostgreSQL <=365 days (364 + 1); these maxima are not verified live, and publication is blocked unless both buckets' live lifecycle rules match. These minimum records survive tenant-content deletion; the tenant ID is pseudonymous, not anonymized. |
| Is data encrypted in transit? | Yes — TLS 1.2 or higher on all connections. |
| Is data encrypted at rest? | Yes. Production and per-customer EC2 root volumes use AWS KMS-backed EBS encryption. Encrypted recovery copies are stored in versioned, TLS-only S3 buckets. Channel PII, MFA secrets, and Slack integration tokens are additionally encrypted at the application layer with AES-256-GCM. Meeting Assistant transcripts use a private S3 bucket with SSE-KMS; Zoom OAuth tokens are AES-256-GCM encrypted. |
| Is data used for advertising or sold? | No. We do not sell personal information or use it for advertising. |
Channel PII encryption: on-call phone numbers and email addresses are encrypted at the application layer (AES-256-GCM) in addition to infrastructure-level protections. MFA secrets receive the same treatment.
Subprocessors
We share personal data only with the subprocessors needed to operate the Service — AWS (hosting, database, transcript storage, and email via Amazon SES), Apple and Google (direct push delivery), Twilio (optional voice delivery and legacy SMS where already enabled), Cloudflare (DNS, edge/Pages hosting for the public sites, Email Routing/Email Worker for inbound alert email, and Workers AI inference for Meeting Assistant transcription (/transcribe) and summarization (/summarize) and the docs assistant; core api.opsping.com traffic is DNS-only and not proxied, ai.opsping.com is a Cloudflare Worker that calls Workers AI, and the docs assistant is a Pages Function), Stripe (billing), and Proton AG (human mailboxes for support, legal, security, sales, general inquiries, migration, SMS compliance, and recruiting/applicant correspondence). Organizations may select default (Workers AI summaries only), customer (their own HTTPS AI endpoint for summaries; transcription remains on Workers AI), or both; for each delivered summary OpsPing records the provider on the summary record and writes a meetingbot.summarized audit entry listing the provider(s) whose summaries were delivered. See the full subprocessor list for purposes, data processed, locations, and SOC 2 status. Customers are notified at least 30 days in advance of subprocessor changes, with a 14-day objection window (see our DPA).
Security Measures
- Access control: AWS IAM with least-privilege access; production access limited to authorized personnel.
- Encryption in transit: TLS 1.2+.
- Encryption at rest: AWS KMS-backed EBS encryption on production and customer-silo root volumes; encrypted backup and recovery storage.
- Application-layer encryption: AES-256-GCM applied to channel values, MFA secrets, and Slack integration tokens.
- Penetration testing: No third-party penetration test has been performed. A structured internal security review was completed on 2026-08-09, and its critical and high findings were remediated.
- Vulnerability management: Automated Dependabot alerts; our target is to apply critical patches within 48 hours of availability.
Compliance Status
| Question | Answer |
|---|---|
| Do you have SOC 2? | No, and certification is not planned. Our practices are mapped criterion-by-criterion to the Trust Services Criteria on the SOC 2 page. |
| Is a DPA available? | Yes — a GDPR-style Data Processing Agreement is available, designating OpsPing as processor and the customer as controller. |
| Do you support EU data transfers? | Yes — via EU Commission Standard Contractual Clauses (2021 SCCs), Module Two (controller-to-processor), incorporated into the DPA. Core Service, the Go summary pipeline, and summaries run in AWS us-east-2; Meeting Assistant Fargate/transcript S3 runs in AWS us-east-1 (both US); summarization and transcription inference runs on Cloudflare Workers AI and edge/Pages delivery on Cloudflare's global network (no regional or jurisdictional pin configured; including public-site request metadata); Stripe, LLC processes billing in the US with global onward transfers under its DPA/SCCs. Cloudflare's and Stripe's own data processing terms and SCCs govern their onward processing as subprocessors; push delivery uses United States Google/Apple provider accounts and their global edge networks under the providers' DPAs/SCCs; human mailbox correspondence is hosted by Proton AG in Switzerland, which is covered by the European Commission's adequacy decision for Switzerland, so no further transfer safeguard applies to that hosting; any onward transfer by Proton outside Switzerland, the EU, or another adequacy country relies on Switzerland-, EU-, or UK-approved standard contractual clauses or another lawful transfer mechanism under Section 11 of Proton's Data Processing Agreement (last modified 2026-02-10). |
| GDPR data subject requests? | Access and deletion requests are honored within 30 days of a verified request. |
Incident Response
- Breach notification: Customers are notified of any personal data breach affecting their data within 72 hours of OpsPing becoming aware of it.
- Status page: OpsPing's own service status is published at status.opsping.com/opsping, and per-customer status pages live at status.opsping.com. Service disruptions are also communicated directly to affected customers by email.
Business Continuity & Disaster Recovery
| Question | Answer |
|---|---|
| Infrastructure resilience? | One active production deployment in AWS us-east-2. A stopped, pre-provisioned recovery environment in us-east-1 can be promoted by an operator; there is no automatic failover. |
| RPO / RTO | Database write-ahead logs are uploaded to us-east-1 every minute, with a daily base backup and nightly logical dumps. Operational targets: RPO ≤ 10 minutes and operator-to-healthy RTO ≤ 90 minutes. These are targets, not contractual commitments. |
| Formal DR plan? | Yes. A documented, operator-led regional recovery runbook exists and has been drilled with a restore. Recovery is not automatic. Email delivery from the recovery region is not enabled. |
Security questions not covered here: security@opsping.com. This questionnaire reflects the current state of the product and will be updated as controls mature.