Security & trust
Your data is yours. We mean that literally.
Crewly reads from Slack, talks to language models, and remembers what your team tells it. Here is exactly how we handle each of those, what we promise, and what we have not built yet.
No model training on your data
We call Anthropic and Voyage APIs with no-train terms in place. Your messages and memory are never used to train a model — ours or theirs. This is contractual, not a setting you have to find.
Encrypted Slack tokens
Your workspace access token is encrypted at rest, decrypted only in memory for the request that needs it, and never written to logs, traces, or analytics. If our logs leak, your token does not.
Tenant isolation by default
Every row of every table is scoped to your organization with Postgres row-level security — enforced at the database, not just the app layer. One team can never read another team's memory, even through a bug.
Crewly only speaks when spoken to
It responds to @mentions and the channels you pick. It does not read channels it was not invited to, and it does not act on its own. You choose where it lives, and you can remove it in one click.