Security & Trust

Last updated: 11/3/2026.

Marcora (MarketCore LLC) gives go-to-market teams governed, on-brand context infrastructure for AI content. This page documents how we protect your data, exactly what is and isn't independently certified today, and what's on our roadmap: the answers a security review needs, stated plainly. Anything not covered here: security@marcora.ai.

Infrastructure and inherited compliance

Marcora runs on a small, deliberately chosen stack of independently audited providers.

ProviderRoleTheir certificationsWhat Marcora inherits
NetlifyWeb application hosting / CDNSOC 2 Type 2, ISO 27001, ISO 27018Edge security, build pipeline, DDoS protection
Anthropic AI content generation, the Marcora agent, and text extraction from uploaded images and some document types (API) SOC 2 Type 2, ISO 27001:2022, ISO 42001 No training on your data under its commercial terms. Marcora agent conversations are stored with Anthropic until we delete them.
OpenRouter Routes requests to OpenAI models, Anthropic (Claude) models and a TypeSafe model for Marcora's AI features, including content generation, the AI Assistant, Context Intelligence, embeddings, extracted document text, lifecycle email, Marcora agent conversation titles, text the Marcora agent sends to Marcora's tools, and memory Not independently assessed by us—
StripePaymentsPCI DSS Level 1, SOC 2We never store card data
Railway Application backend, database, file delivery, document conversion, and integration services (US East region) SOC 2 Type 2, SOC 3 Physical, network, and infrastructure controls; encryption at rest; audited platform operations
Mailgun (Sinch)Transactional emailSOC 2 Type 1 & 2, ISO 27001Audited email delivery
ComposioThird-party integrations layerSOC 2 Type 2, ISO 27001:2022Audited integration handling
PostHog Product analytics, web-app session recordings and error reports (without passwords, payment details, secrets, saved memories, or conversations with the Marcora agent and AI Assistant), and usage metrics from AI requests (without prompt or answer text) SOC 2 Type 2Audited analytics processing

What's independently certified, and what isn't: the certifications above are held by our infrastructure providers, and Marcora inherits the physical, network, and infrastructure controls they cover. Marcora (MarketCore LLC) has not yet completed its own independent SOC 2 or ISO 27001 audit. Our compliance roadmap is below, and enterprise customers can request our security documentation package at security@marcora.ai.

Additional subprocessors

These providers also process data on our behalf. They are listed separately from the table above because we disclose here what each one receives rather than certifications we have independently assessed.

  • OpenRouter — requests to OpenAI models, Anthropic (Claude) models and a TypeSafe model from Marcora's AI features, including content generation, the AI Assistant, Context Intelligence, embeddings, text extracted from your uploads, lifecycle email, Marcora agent conversation titles, text the Marcora agent sends to Marcora's tools, and memory (the messages you write and your saved memories).
  • RapidAPI — receives the raw bytes of PDFs you upload, for text extraction.
  • Affonso — receives every user's name and email address at signup, not only those arriving through a referral.
  • Perplexity, Exa and Brave — receive the research and search queries you run. Exa and Brave also receive the web searches the Marcora agent runs through them.
  • Browserless and URLtoText — receive the URLs Marcora uses them to fetch, including for the Marcora agent.
  • Google Cloud Storage — holds database backups and stored file objects.
  • Braintrust and Langfuse — receive a 0.01% sample of request traces as usage metadata, tagged with our internal user, team and session IDs. Privacy mode is enabled for both, so prompts and completions are stripped before the trace is sent.

Encryption

  • In transit: TLS 1.2 or higher on every connection to Marcora over the internet. Traffic between our own services inside Railway's private network, such as from our backend to our file-storage service, is encrypted by that network (WireGuard) rather than by TLS.
  • At rest: provided by our infrastructure providers, plus application-level AES-256-GCM encryption (per-file random IVs, authenticated encryption, dedicated key management) for files in our file-delivery service. The credentials we store for each Slack, Google Chat and Buzz workspace use application-level AES-256-CBC encryption; Microsoft Teams needs no stored credential.

Your data and AI

  • You own your content. We process it only to provide the Service.
  • Marcora and its AI model providers don't train on your content. We call Anthropic directly under its commercial API terms, which don't allow it to train models on that content. Requests to OpenAI models and a TypeSafe model, and some requests to Anthropic's Claude models, are routed through OpenRouter, whose data-collection settings on our account are configured so that neither paid nor free endpoints train on request data and prompts are never published. Every request we send also tells OpenRouter to use only providers that don't collect request data.
  • Data residency: Marcora's application backend and database are hosted in the United States, on Railway's own infrastructure in its US East region. Google Cloud Storage holds database backups and stored file objects in the US. AI requests aren't pinned to a region (see Section 8 of our Privacy Policy).

Access controls

  • Workspace role-based access control: admin, creator, and collaborator roles.
  • Least-privilege access to production systems; multi-factor authentication on all administrative and infrastructure accounts.
  • Programmatic access (API and MCP server) is authorized via OAuth 2.0 or workspace-scoped API keys that customers manage and can revoke at any time.
  • Signed, expiring URLs (15-minute validity) for file downloads.

Data retention and deletion

  • Cancelling a subscription downgrades you to our free tier; your data stays yours and stays put.
  • Account deletion (requested by emailing support@marcora.ai) is completed in our live systems within 30 days, except where law requires retention. Backup copies expire within 100 days. For Marcora in Slack, Google Chat, Microsoft Teams and Buzz, deletion also covers the conversation records Anthropic keeps for us, and chat messages copied into our request logs are removed on their own 7-day schedule (see Section 2.4 of our Privacy Policy).
  • Removing Marcora from a Slack workspace deletes that workspace's Marcora conversations from our live systems within 14 business days. Backups and some other copies aren't covered (see Section 2.4 of our Privacy Policy).
  • Your data is exportable at any time; you are never locked in.

Business continuity

  • Automated backups via our infrastructure providers; data exportable at any time in standard formats.
  • For enterprise agreements we offer continuity assurances on request, including data-export guarantees and escrow arrangements.

Responsible disclosure

Found a vulnerability? Email security@marcora.ai. We acknowledge reports within 2 business days.

Compliance roadmap

  • Today: independently audited (SOC 2 Type 2 / ISO 27001) infrastructure across every provider in our stack; application-level file encryption; DPA available on request; this page.
  • Next: documented internal security policies (access, incident response, retention); subprocessor-change notifications.
  • Planned: independent application-layer penetration testing; SSO and MFA for customer accounts; SOC 2 certification for MarketCore LLC.

Enterprise prospects: request our security documentation package at security@marcora.ai. It includes our security overview and DPA, plus direct paths to our infrastructure providers' own trust portals (Railway, Netlify, PostHog, and others), where their attestation reports are available first-hand.