Security
Last updated: 11 July 2026. How truelo protects your data.
Your code never leaves your systems
truelo never fetches, stores or transmits your source code, diffs or file contents. From GitHub we read only commit metadata — subject line, lines changed, repo, branch, timestamps — and from Jira/Linear only issue metadata. Anything that looks like a secret (API keys, tokens, passwords) is stripped from commit messages and PR titles before it is stored or sent to the AI provider.
Per-workspace isolation
Every workspace's data is scoped to its own organization. Queries are org-scoped server-side, so one workspace can never read or modify another's data — a property we verify with automated cross-tenant tests. The AI only ever sees one person's day at a time.
Encryption
All traffic is served over TLS (HTTPS). Data is encrypted at rest by our database provider, and integration credentials (GitHub/Slack/Jira/Linear tokens and any bring-your-own AI key) are additionally encrypted at the application layer before they are stored. Server-side credentials are never exposed to the browser.
Access & authentication
Sign-in is via Slack — there are no passwords for us to store. Access inside a workspace follows the roles your admins assign (owner / admin / member / viewer). Every API route and page enforces authentication and the appropriate role, and fails closed if configuration is missing — an invariant covered by our automated test suite.
Data minimisation
We collect only what the product needs: identities, work metadata, and what people choose to submit via /wrap and time-off requests. We do not use advertising or third-party analytics trackers. See the Privacy Policy for the full list.
AI processing
Summaries and ratings are generated by a third-party AI provider (OpenAI or Anthropic); the metadata and notes sent are processed to produce the output and are not used to train their models under the applicable API terms. Untrusted text is passed to the model as clearly-delimited data, not instructions. Workspaces may bring their own AI key so processing runs on their own provider account.
AI outputs are advisory decision-support, not automated decision-making: any per-person rating or summary is a signal for a human to review and is never the sole basis for an employment decision. There is always a human in the loop. An admin can switch off per-person AI ratings and summaries entirely in Organization settings — dashboards, effort estimates and delivery metrics continue to work without any AI evaluation of individuals.