Agents
Sign in to your account
Agents
Setup Required
Complete the configuration to start using the assistant
No data sources connected
Use + Add Connection above to get started
All connections are read-only. We never write to your database. Passwords are encrypted at rest.
Add Data Source
Your database admin needs to whitelist our IP to allow connections.
Default is usually fine. Change only if your tables live in a different schema.
Supported: .xlsx, .xls, .csv (max 50 MB)
Share the sheet with (loading service account...) as Viewer first.
Each tab becomes a separate table. You can refresh later to pull updated data.
Coming soon
Lorem ipsum dolor sit amet, consectetur adipiscing elit. Sed do eiusmod tempor incididunt ut labore et dolore magna aliqua. Ut enim ad minim veniam, quis nostrud exercitation ullamco laboris nisi ut aliquip ex ea commodo consequat.
Coming soon
Tell us about your business. Edit the template below or replace it with your own text. This helps the system understand what your data means.
Choose which tables to include. Uncheck tables you don't need (unused, logs, or noisy) to keep the system focused.
Document each included table and its columns. The system uses this to understand your data. Click Suggest with AI to have Claude propose a description based on your business context and sample rows — you edit before saving.
Select a table on the left to document it.
Capture domain knowledge per table — codes, special states, classification rules. The AI uses this when generating preprocessing SQL. Manual textarea, or let Claude propose candidates.
Select a table on the left to capture business rules.
Write SQL that cleans or normalizes source tables before they're used downstream. Each preprocessing query produces a silver table named <connection>_<name> (e.g. main_pos_ventas) that the Queries tab reads from.
Select a preprocessing query on the left, or click + New preprocessing query.
This step is coming next.
We'll wire the next phase in the next iteration.
No active connection
Go to Data and activate a connection to continue.
Invite a teammate
We'll email them a link to set their password and join your account.
Preview
No tabs yet
Create a tab first, then add metrics to it. Tabs let you organize your dashboard by topic or audience.
Configure Dashboard
Tabs
Add metric
Declare relationships between any two materialized tables. Joins span connections — pair a Google Sheets dictionary with a SQL Server fact table, etc.
Pending AI proposals
Accept to add them to your saved joins; reject removes them and the AI won't propose them again.
Saved joins
Materialized tables
Reference list — the universe joins can draw from.
Query inspector
Run a read-only SELECT against the warehouse. Reference any silver table by its short name (e.g. main_pos_ventas). Returns up to 20 rows.
Add join
Pick the two tables and the columns that match. Use the Test button to preview rows before saving.
Build the gold layer: queries that read silver tables, join them with the relationships above, and materialize as mart_<name> for the reporting step to consume.
Three ways to author: + New datamart writes SQL manually · type a name + description in the editor and click Generate with AI for AI-assisted drafting · or click Suggest with AI to see fully proposed candidates.
Select a datamart on the left, or click + New datamart.
AI datamart suggestions
Click Use this to load a suggestion into the editor — you can edit before saving.
Build the final layer: queries that read your datamarts and produce the answers dashboards consume. Materialized as report_<name>.
Three ways to author: + New report writes SQL manually · type a name + description and click Generate with AI for AI-assisted drafting · or click Suggest with AI for fully proposed candidates.
Select a report on the left, or click + New report.
AI report suggestions
Click Use this to load a suggestion into the editor — you can edit before saving.
No materialized tables yet
Go to Knowledge → Preprocessing and run a pipeline first.
No workflows yet
Create an automated email report. Pick data, optionally add an AI summary, send.
Choose workflow type
Pick what this workflow should do. You'll configure the details on the next step.
Pull data from a source, optionally add an AI analysis paragraph, and email an HTML report to one or more recipients.
Slack notifications, scheduled CSV export, dashboard refresh — and more.
New workflow Email sender
Up to 7 cards per workflow. They render in the email in the order you add them.
Recent runs
No agents yet
Create a specialized AI assistant. Give it a role and a context — your team can chat with it.
Every agent reads only your published Reports — never silver, mart, or the raw source database. Per-agent token limits protect your monthly cost.
New agent
Shown next to the agent in the chat picker.
Global safety rules and the role's data-only behavior always apply, even if you don't mention them here.
Advanced
Stops the agent if it loops on tool calls. 8 is a safe default.
Leave empty to use the account-wide daily limit.
Execution history across the platform — pipeline chains, scheduled runs, workflows, and agents. Each row says what happened, and why if it failed.
Email preview
This is exactly what recipients will receive.
Account
Loading…