Tools give agents access to external capabilities. Use tools when an agent needs to search documents, call APIs, query systems, or use an MCP server.

Tool areas

Pinter separates tool-related work into these app areas:

Available tool types

Agents can use two built-in ADK tools and several configurable tool types: Configurable types store their credentials and settings as tool instances. Assign the tool instance to an agent in the builder; the backend resolves the reference at run time. See Backend and ADK architecture for the reference format.

In-memory Tabular Data Querying

When structured files (.csv, .tsv, .xlsx) are uploaded to Knowledge Hub, Pinter automatically profiles their schemas:
  1. Schema Profiling: Identifies column names, inferred types, and row count limits.
  2. Dynamic In-memory SQLite DB: Instead of forcing thousands of rows of raw CSV into the model context window (which wastes tokens and causes hallucination on aggregation), Pinter stages the spreadsheet into an in-memory database.
  3. Automatic Tools:
    • get_table_schema(document_id): Agent inspects table column names, datatypes, and sample rows.
    • query_tabular_data(document_id, sql_query): Agent executes SQL aggregate queries (SUM, AVG, GROUP BY, WHERE filters) directly over the tabular data.
1

Create or connect the tool

Add the tool in the relevant app area.
2

Test the tool

Verify the tool works before assigning it to an agent.
3

Assign the tool to an agent

Open the agent builder and add the tool to the agent configuration.
4

Run the agent

Use chat or evaluation to check the result.
Only connect tools that the agent is allowed to use. Tools can expose sensitive data or perform external actions.