Skip to content

Account and security

Your data and privacy

What SentientOne stores, how long it keeps it, what's redacted in traces, how secrets are encrypted and which services process data.

Concept2 min readUpdated 23 October 2026

This page explains, in plain terms, what happens to your data. The legal version is in the Privacy policy; security details are on the Security page.

What's stored, and for how long

DataKept
Agents, prompts, versions, workflowsUntil you delete them or your account
Knowledge: documents, crawled pages, chunks and their embeddingsUntil you delete the source, the agent or your account
Conversations and messagesUntil you delete the agent or your account
Traces (requests, spans, redacted prompts and replies)90 days
Billing history and receiptsWith your account; Stripe keeps its own records

What traces capture

Traces record model inputs and outputs and each tool call's input and output so you can debug answers. Before they're stored, email addresses, API keys and phone numbers are redacted, and each value is cut to 10 KB.

Secrets

  • Model provider keys, MCP server credentials, workflow signing secrets and Slack webhook URLs are encrypted with AES-256-GCM before they're stored, and are never sent back to your browser in full.
  • Webhook trigger tokens are stored only as a SHA-256 hash, and widget visitors' IP addresses are hashed.
  • Two-step verification secrets are stored encrypted.

Services that process data

ServiceWhat it receives
Your model provider (OpenAI, Anthropic, Google or Groq)The prompt for each reply: system prompt, history, retrieved passages and tool results, on your own key
OpenAI (embeddings)Knowledge text when it's added, and questions, to make embeddings
CohereThe question and up to 24 candidate passages, to rerank them
FirecrawlThe website addresses you ask to crawl
SupabaseThe database and file storage
ClickHouseTraces, with the redactions above
UpstashRate limits, caches and recent conversation history
StripePayments and cards
ResendAccount, billing and workflow emails
PostHogProduct analytics: ids, counts and timings only, never message text or keys
Google Cloud and VercelHosting for the API and the app

Export

There's no self-serve export. Conversations are readable on each agent's Conversations tab, and traces through the API (GET /v1/traces). For anything else, email support@sentientone.ai.

Is my data used to train models?

SentientOne doesn't train models. Your model provider handles prompts under your own account's terms with them.

Where do I report a security issue?

Email security@sentientone.ai with the details.

Still stuck?

Email us with what you tried and any request id. A person replies.

support@sentientone.ai