Skip to main content
Knoq is designed with a deliberately small data footprint. The platform stores only what is necessary to render your session history, manage your integrations, and operate the service — nothing more. This page describes exactly what is written to Knoq’s database, what is never stored, how long data is kept, and how you can delete it.
Knoq fetches from your connected tools only at the moment you ask a question. There is no background sync, no continuous crawl, and no indexed copy of your Slack workspace, Notion pages, or GitHub repositories. Every tool call happens on demand, and only the metadata required to render your transcript is persisted afterward.

What Knoq stores

Account data

When you sign in via your identity provider (Google or enterprise SSO), Knoq stores your email address, display name, and organization membership. This is the minimum required to authenticate you, enforce seat limits, and send transactional emails (invite notifications, billing receipts, trial expiry warnings).

Session data

Every chat session you create is stored in Knoq’s hosted database, along with the event log that makes up the transcript: your question, the AI’s responses, tool call events, status transitions, and usage metadata (token counts and cost figures for billing). This data is what allows you to return to a session after closing your browser, and it allows multiple tabs or team members to view the same session concurrently. Session content — your question and the agent’s answer — is stored in plaintext in the event log, scoped strictly to your user and organization. No member of another organization can access your sessions through any API endpoint.

Integration credentials

When you connect a tool such as Slack, Notion, GitHub, or Linear, Knoq stores the resulting OAuth access token (and refresh token where applicable). Every credential is encrypted with AES-256-GCM before it is written to the database. The plaintext token is held in memory only for the duration of a tool call and is never written to logs, returned through any API response, or accessible outside of the internal request handling layer.

Agent memory

If agent memory is enabled for your organization, Knoq stores a set of key-value pairs derived from your sessions — preferences, recurring context, and similar signals — to improve the quality of answers over time. Memory entries are encrypted, scoped strictly to your organization and user, and can be cleared on request.

Audit logs (Business)

On the Business tier, Knoq maintains an append-only audit log of administrative actions: member invites and removals, role changes, SSO configuration updates, settings changes, and similar events. Audit rows are never updated or deleted — they form a tamper-evident record of your organization’s administrative history.

Verified answers

Answers that your team has marked as canonical (“verified”) are stored so Knoq’s AI pipeline can inject them as high-confidence context on future queries. These are accessible to all members of your organization.

What Knoq does NOT store

The following data never enters Knoq’s database:
  • Full document bodies from your connected tools — Notion page contents, Slack thread bodies, GitHub file contents, Linear issue descriptions, and similar source data are fetched on demand and used only to construct the agent’s answer. They are not copied, indexed, or cached beyond the current request.
  • Persistent tool-result content — tool call outputs are included in the session event log only as metadata. Payloads above a configured size threshold are automatically redacted before being written to the database.
  • Prompt contents sent to AI providers — while Knoq constructs and sends prompts to your AI provider on your behalf, the full prompt is not stored separately. The session event log contains the rendered transcript, not the raw model input.
  • Payment card data — full payment credentials are handled exclusively by Knoq’s Merchant of Record and are never transmitted to or stored by Knoq.
Because Knoq does not index your source tools, permission changes in those tools take effect immediately on the next query. If a Notion page is restricted or a Slack channel is made private, Knoq can no longer read it the moment the permission changes — there is no stale cached copy to worry about.

Data retention

Data classRetention
Session history and event logsRetained while your organization is active, until you delete them
Integration credentials (OAuth tokens)Deleted immediately when you disconnect the integration or close your account
Agent memoryRetained until cleared by you or until your account is deleted
Account and usage dataRetained while your organization is active; deleted upon organization deletion
Audit logsRetained for compliance purposes — contact support@knoq.one for your data retention SLA
Billing recordsRetained as required by applicable tax law — contact support@knoq.one for details
Database backupsManaged by Knoq’s cloud infrastructure provider — contact support@knoq.one for your backup retention SLA

Deleting your data

Deleting sessions

You can delete individual sessions from the sidebar in the chat interface by hovering over a session and selecting Delete. To delete all sessions at once, go to Settings → Sessions and select Delete all sessions. Deletion is immediate and permanent — the session row and all associated events are removed from the database.

Disconnecting integrations

To remove an OAuth token for a connected tool, go to Settings → Integrations, find the integration you want to remove, and click Disconnect. The encrypted token is deleted from Knoq’s database and the OAuth grant is revoked in the source tool.

Deleting your account

To request full account deletion — including your profile, organization membership, agent memory, and any remaining data — contact support@knoq.one. We will complete the deletion within 30 days and confirm by email. Audit logs and billing records may be retained longer as described in the retention table above.

PII handling

Knoq’s AI pipeline automatically inspects prompts before they are sent to an AI model provider and strips personally identifiable information (PII). This reduces the risk of sensitive data such as email addresses, phone numbers, or identification numbers appearing in model inputs that transit Knoq’s infrastructure or your BYOK provider’s API. Additionally, tool-result payloads that exceed a configured size threshold are redacted before being written to the event log. This prevents inadvertently large document chunks from being stored in the transcript database.
PII redaction is a best-effort defense-in-depth control, not a guarantee. You should not rely on it as a substitute for limiting the scopes granted to your integration tokens, or for configuring your connected tools’ own data access policies.

Data residency

Knoq is hosted on Knoq’s cloud infrastructure, primarily in the United States. Session data, audit logs, and credentials are stored in Knoq’s hosted database with in-region data storage. If your organization requires data residency in a specific region or jurisdiction, contact security@knoq.one to discuss your requirements. Enterprise MSA customers can request a Data Processing Agreement (DPA) and our current subprocessor register.