generated: '2026-08-12'
method: searched
source: https://docs.ploy.ai/security
docs:
- https://docs.ploy.ai/security
- https://ploy.ai/pricing
- https://trust.ploy.ai
standards:
- id: oauth2
  conforms: true
  role: client
  evidence: >-
    "Connections use OAuth 2.0 with scoped permissions" for every third-party
    integration (Google, GitHub, Figma, HubSpot, Notion, Slack, ad platforms);
    "Ploy never stores passwords for any integration". Ploy is an OAuth CLIENT
    only — it publishes no authorization server of its own
    (/.well-known/oauth-authorization-server 404, probed 2026-08-12).
  source: https://docs.ploy.ai/integrations
- id: oidc
  conforms: false
  evidence: /.well-known/openid-configuration returns 404 on ploy.ai (probed 2026-08-12)
- id: rfc9457-problem-details
  conforms: false
  evidence: >-
    Errors use a bare {"error": "..."} JSON envelope, not
    application/problem+json (observed live 2026-08-12).
- id: rfc9116-security-txt
  conforms: true
  evidence: https://ploy.ai/.well-known/security.txt returns 200 with a Contact field
- id: rfc8594-sunset-header
  conforms: false
  evidence: no deprecation or sunset policy published
- id: bearer-token-auth
  conforms: true
  evidence: >-
    Both credential types (PLOY_API_TOKEN, webhook endpoint keys) are sent as
    HTTP bearer tokens in the Authorization header.
- id: openapi
  conforms: false
  evidence: >-
    No OpenAPI/Swagger document found on ploy.ai, docs.ploy.ai or ingest.ploy.ai
    after probing /openapi.json, /openapi.yaml, /swagger.json, /v1/openapi.json,
    /api-docs, /redoc, /docs.json (2026-08-12).
- id: asyncapi
  conforms: false
  evidence: webhook ingest is documented in prose; no AsyncAPI document published
- id: mcp
  conforms: false
  evidence: >-
    No MCP server found. No mention of MCP or Model Context Protocol anywhere in
    ploy.ai/llms.txt, docs.ploy.ai/llms.txt or the docs pages; /mcp and
    /.well-known/mcp.json 404 (probed 2026-08-12). Ploy's agent-facing strategy
    is the CLI plus its own installable Agent Skills catalog.
- id: a2a
  conforms: false
  evidence: >-
    /.well-known/agent-card.json and /.well-known/agent.json return 404 on both
    ploy.ai and docs.ploy.ai (probed 2026-08-12).
- id: agent-skills
  conforms: true
  evidence: >-
    Ploy publishes an installable Agent Skills catalog — `ploy skills init`
    writes .agents/skills/<id>/SKILL.md and .claude/skills/<id>/SKILL.md into a
    checked-out Ploy site, with ownership tracked in .ploy/skills.json and drift
    detection via `ploy skills sync --check`.
  source: https://docs.ploy.ai/cli/local-development
- id: do-not-track
  conforms: true
  evidence: install.sh honors the DO_NOT_TRACK convention (and PLOY_INSTALL_NO_ANALYTICS)
  source: https://ploy.ai/install.sh
- id: llms-txt
  conforms: true
  evidence: llms.txt served at both https://ploy.ai/llms.txt and https://docs.ploy.ai/llms.txt (HTTP 200)
compliance:
- id: soc2-type-ii
  status: published
  evidence: >-
    Listed as an Enterprise plan entitlement on https://ploy.ai/pricing and
    surfaced through the Vanta-hosted trust center at https://trust.ploy.ai
    ("runploy.com Trust Center", HTTP 200, probed 2026-08-12). The certification
    report itself is behind the trust center's access request.
- id: gdpr
  status: not-published
  evidence: privacy policy exists (https://ploy.ai/privacy) but no GDPR/DPA page was found
data_handling:
  ai_training: >-
    "Ploy does not use your content to train AI models"; Enterprise terms
    restate this as "no training on your data".
  encryption_at_rest: AES-256
  encryption_in_transit: HTTPS/TLS for all sites, ACME-provisioned certificates per custom domain
  build_isolation: builds run sandboxed and cannot access other workspaces or sites
  deploys: immutable snapshots with instant rollback
  visitor_privacy: cookieless visitor identification, privacy-first first-party analytics
  source: https://docs.ploy.ai/security