10 min
Connect a database
Give a workflow Postgres, MySQL, or ClickHouse. It reads the schema, writes real SQL, and checks the results in its own sandbox.
5,000 free credits to start.
- 01
Create a read-only role first
Before you touch Helios, create a database role that can read the schemas a task needs and nothing else. Your database enforces the rule, which is where that decision belongs.
- 02
Add the connection
From Settings, then Integrations, add a Postgres, MySQL, or ClickHouse connection with a host, port, database, and credentials. The connection string is encrypted and decrypted only when a run needs it.
- 03
Reach a private database through an SSH tunnel
When the database sits on a private network, add an SSH tunnel on the connection. Runs connect through the bastion you name. Nothing is exposed to the public internet.
- 04
Ask a question you already know the answer to
Start a chat, attach the database, and ask something you can check. The agent reads tables, columns, types, and keys, writes the SQL, runs it, and explains the result. Compare it with the answer you knew.
- 05
Read the SQL it wrote
Every run records the queries the agent ran and the rows they returned. When a number looks wrong, the query that produced it is in the transcript.
- 06
Reuse the prompt as a workflow
Create a workflow with the same prompt and add a schedule. Put stable facts about the schema in the prompt so every run starts with them.
Keep going
Next guides
7 min
Pin repositories as context
Pin a GitHub repository to a workflow or a chat. The agent starts every run with it.
8 min
Put a workflow on a schedule
Turn a working prompt into a standing task. Describe the cadence the way you would say it out loud. Helios writes the cron expression.
8 min
Use secrets and variables
Keep the values a task needs out of the prompt. Variables hold configuration. Secrets hold credentials. The agent never sees a secret in plain text.
Templates
Put it to work
Templates that use what this guide covers.
Operations
Weekly metrics report
Queries your warehouse, computes the metrics your team argues about, and posts a report with the outliers called out.
Operations
Nightly data quality checks
Runs your invariants against production every night and emails the on-call analyst which rows broke and since when.
Sales & RevOps
Churn risk watchlist
Joins billing signals with product usage and support volume to find the accounts drifting toward the exit.
Try it in your workspace.
Start free, connect a tool, and write the first prompt in plain words.