Agent Serve · Data sources
Connect the warehouses and databases you already trust. Agent Serve turns read-only queries into datasets that dashboards, metrics and monitors read.
Agent Serve Data sources connects Databricks, Snowflake, BigQuery, PostgreSQL, MySQL, SQL Server, Salesforce, Qlik and Excel workbooks to Agent Serve, the AI workspace where teams build, deploy, and manage AI agents. Each read-only query becomes a dataset, saved into a workspace table on a schedule or queried live, which dashboards, metrics, rules and monitors use.
How it works
Agent Serve keeps the source where it is and brings in only the rows a query returns.
Save host, user and token as secrets. The connection names the secrets, never the values.
Pick a source from Available sources and run Test connection.
Write a SELECT or WITH query. Agent Serve blocks anything that writes.
Save the result into a table on a schedule, or query it live on each request.
By the numbers
9
Source types you can connect
50,000
Rows per dataset, at most
15 min
Shortest scheduled refresh
10
Refresh versions kept per dataset
Figures are read from the product code.
In depth
Agent Serve turns a source into datasets, and datasets into the tables every analysis module reads.
An Agent Serve dataset is either saved into a workspace table on a schedule, or queried live each time someone asks.
Agent Serve dashboards show which source a number came from and when it last synced.
Agent Serve accepts one SELECT or WITH statement per query and blocks write keywords. The source login’s own permissions remain the final limit.
Connectors
Agent Serve connects Databricks, Snowflake and BigQuery, and runs your SQL where the data lives.
Databricks · Snowflake · BigQuery
Agent Serve connects PostgreSQL, MySQL and SQL Server, with SSL set per connection.
PostgreSQL · SSL require
Agent Serve queries Salesforce with SOQL and reads Qlik Cloud or Qlik Sense Enterprise apps and sheets.
Salesforce · Qlik
Agent Serve reads Excel workbooks and CSV files from the Library, with a sheet and range picker.
Loan tape.xlsx · Sheet1!A1:K
Examples
Three datasets financial services teams build in Agent Serve.
An Agent Serve dataset pulls the loan book from Snowflake each morning into a table that the arrears dashboard reads.
An Agent Serve dataset reads open claims from PostgreSQL every hour, and a monitor watches the backlog.
Finance uploads the budget workbook to the Library, and the Agent Serve dataset refreshes whenever the file changes.
Examples show what teams can build in Agent Serve. They describe patterns, not named customer deployments.
Before and after
| Work | Before | With Agent Serve |
|---|---|---|
| Getting data in | BeforeSomeone exports a CSV each Monday and emails it round. | With Agent ServeAn Agent Serve dataset refreshes on a schedule from the source. |
| Database passwords | BeforePasted into each report’s connection settings. | With Agent ServeKept once as secrets, named by the connection, decrypted only at query time. |
| Knowing what changed | BeforeCompare two spreadsheets by eye. | With Agent ServeVersion history lists the rows and columns each refresh changed. |
Under the hood
| Sources | Databricks, Snowflake, BigQuery, PostgreSQL, MySQL, SQL Server, Salesforce, Qlik and Excel or CSV. |
|---|---|
| Where it lives | The Data sources tab in Team spaces. |
| Credentials | Named secrets in Settings, decrypted only when a query runs. |
| Queries | One SELECT or WITH statement, up to 20,000 characters, with write keywords blocked. |
| Datasets | Materialized into a workspace table on a schedule, or live per request. |
| Refresh | Manual, every 15 minutes, hourly, every 6 hours or daily. |
| Size | Up to 50,000 rows per dataset, or a lower cap you set. |
| Usage | Refreshes and live queries are recorded in Settings › Usage. |
Governance
Agent Serve reads from sources with the permissions you give it, and keeps the keys out of the connection itself.
Security and governanceAgent Serve stores source credentials as encrypted workspace secrets, and a connection only names them.
Each Agent Serve data source has its own access level on top of the team space roles.
Admin, write and read roles per workspace, plus permission groups that limit which models, blocks and integrations a team can use.
Row-level policies narrow what each person sees in tables, dashboards, exports and scheduled reports.
Run Agent Serve in your own cloud or data centre, next to the systems it reads.
Works with
Materialized datasets land as workspace tables.
Panels read datasets and show their source and freshness.
Define governed metrics over the tables a dataset fills.
Describe dataset tables as business entities.
Excel and CSV files in the Library become datasets.
Further reading
Accruals live in the gap between what was received and what was invoiced. Here is a setup where three workflows build the accrual list, a controller approves it, and next month’s true-up runs on its own.
Bank finance · 13 min readThe morning spreadsheet is out of date before the meeting starts, and two people bring two RTO numbers. This playbook moves daily decisions onto dashboards that refresh every 15 minutes or query live, with one definition per number.
Ecommerce leadership · 5 min readPricing decisions often rest on a spreadsheet only one analyst fully understands. Here is a scenario model built on a dashboard, where every assumption is visible and every scenario sits next to the others.
Lending · 5 min readQuestions
Agent Serve connects Databricks, Snowflake, BigQuery, PostgreSQL, MySQL, SQL Server, Salesforce, Qlik, and Excel or CSV files from the Library.
No. Agent Serve accepts one SELECT or WITH statement per query and blocks write keywords. Give the source login read-only permissions as well.
Agent Serve Data sources bring rows in for analysis as datasets. Integrations give agents and workflows actions and triggers in business apps.
Not today. Agent Serve datasets fill workspace tables, and dashboards, metrics and monitors read those tables.
Last updated
Enterprise access
Agent Serve is available to teams by invitation. Tell us what you want to build and we will set up your workspace with you.