Snowflake
How to connect Snowflake read-only for AI analytics
The exelens team · 1 min read · Last updated
In short
Create a dedicated service user on an RSA key pair, give it a reader role with SELECT on views only, put it on a small warehouse with a resource monitor, and restrict it with a network policy. That is enough for exelens to read safely.
The checklist
- A dedicated service user. Never reuse a person's login.
- An RSA key pair. Put the public key on the user and keep the private key for the tool. See Snowflake's key-pair guide.
- A reader role. Grant
SELECTon the views people may ask about, not on tables. - A small warehouse. Create a dedicated one with a resource monitor so a runaway question cannot burn credits.
- A network policy. Allow only the tool's egress addresses on the service user.
Let the script do it
In exelens, Data → Connect → Snowflake → Show read-only setup script produces the statements for steps 3 and 4. An administrator runs them once. Then paste the account identifier, warehouse, database, role, user and private key, and press Test connection. exelens warns if the role has write privileges.
Why views
Granting access to views keeps the AI on definitions your team already agreed. See why views beat tables, and the full list on exelens for Snowflake. Book a demo if you want us to walk through it with you.
Frequently asked questions
- Can I use a password instead of a key pair?
- You can if you must, but a key pair removes the shared secret and is the preferred option.
See it on your own warehouse
A short walkthrough with the founder: connect read-only, ask a real question, pin the answer.