Security
An assistant you can give real access to
Indexbrain is only useful if it can reach the tools your work lives in, which means the question is not whether it has access but what it is allowed to do with it. Every answer below is a property of the product, not a policy we intend to adopt.
Access
What it may do
Read-only on connect
Writing is off by default at the organisation level. A tool you connect today can be read and nothing else, and no configuration step turns that into write access by accident.
Per action, not per app
Each individual action a tool exposes is set to ask, always allow, or disabled. Indexbrain can be allowed to comment on an issue without being allowed to close one.
Every write waits for a person
Anything that changes a system outside your workspace stops and asks. There is no auto-approve mode anywhere in the product — not a hidden one, not one behind a plan.
Scheduled work can never hold the keys
The highest-risk tools are disqualified from unattended runs entirely, on every plan and with any setting. A cron job cannot inherit shell access.
Your data
What we keep, and for how long
Raw content is not kept
Source material is purged on a 90-day cycle, and the transcripts from AI-client syncs are dropped as soon as they have been read. What persists is the derived, searchable layer — not your inbox.
Nothing trains a model
Your content is not used to train anything, ours or a vendor's. It is sent to a model to answer your question and that is the end of it.
Encrypted in transit and at rest
HTTPS everywhere, and the credentials that matter most — OAuth tokens and API keys — are encrypted at rest with Fernet rather than stored as text.
One organisation cannot see another
Every query in the product is scoped by organisation. It is not a filter applied at the end; it is a column on every table and a condition on every read.
Being straight with you
What we don’t claim
We hold no security certifications today. No SOC 2, no ISO 27001, no HIPAA attestation, and no completed third-party penetration test. Plenty of sites in this category print those seals in the footer. A certification is a claim about an audit that either happened or it did not, so ours stays empty until one has. When that changes, this page will say so and name the report.
Per-object permissions are real for some sources, not all. For eight of the connected tools — including Slack private channels, Google Calendar, Drive and the meeting recorders — Indexbrain resolves who may see each individual item and filters on it. For the rest, content is visible to everyone in your organisation, because those APIs do not expose per-item permissions to an integration. That is a boundary we cannot cross, and the honest answer is to tell you which tools are on which side rather than describe the whole product as permission-aware.
We are early. Indexbrain is a young product. The design decisions above are deliberate and they are in the code, but they have not been through an external audit, and you should weigh that alongside everything else here.
Found something?
If you believe you have found a vulnerability, write to security@indexbrain.online. Tell us what you found and how to reproduce it, and give us a reasonable window to fix it before making it public. We will confirm we have received it.
For how we handle personal data, see the Privacy Policy and the Cookie Policy.
Hand it the first job this week
Add it to Slack, connect the tools it may read, and ask for something you would otherwise do yourself.
Join early accessFree to start · Nothing goes out without your yes
Email to Northwind