Skip to main content

Built-in Tools

Roundtable gives the AI access to 16 built-in tools that let it take real actions — searching the web, running code, querying databases, managing files, and more. Tools are what turn a chat model into a productive workspace assistant.

Tools are organized into six categories:

CategoryToolsDefault
Web Toolsweb_search, read_urlEnabled
Code Toolsrun_code, shell_exec, calculatorEnabled
File Toolsread_file, write_file, list_files, find_fileEnabled
Git Toolsgit_clone, git_commitEnabled
Data Toolsquery_bigquery, query_snowflake, query_databricksEnabled
System Toolsdescribe_workspace, bridge_workspaceAlways enabled

Web Tools

Search the web using Google Search and return summarized results.

When it's useful: Answering questions about current events, finding documentation, researching topics the AI doesn't have training data for.

Example prompt:

What are the latest changes in the Snowflake pricing model?

read_url

Fetch and read the content of a specific URL, returning the page text.

When it's useful: Reading documentation pages, extracting data from public APIs, reviewing content at a known link.

Example prompt:

Read the README from https://github.com/example/repo and summarize the setup steps.

Code Tools

run_code

Execute code in a sandboxed environment. Supports Python, JavaScript, and other common languages.

When it's useful: Data transformations, quick calculations, prototyping logic, validating ideas with real execution.

Example prompt:

Write a Python script that reads a CSV and calculates the average of the "revenue" column.

shell_exec

Run shell commands directly in the workspace environment.

When it's useful: Installing packages, running build scripts, checking system state, interacting with CLIs.

Example prompt:

Install pandas and matplotlib, then check which Python version is available.
warning

shell_exec runs commands in the workspace's isolated pod. It has full access to the workspace filesystem but is sandboxed from other workspaces and the host system.


calculator

Evaluate mathematical expressions and return precise numeric results.

When it's useful: Quick arithmetic, unit conversions, financial calculations where floating-point precision matters.

Example prompt:

What is 15% of $4,250 compounded monthly over 3 years?

File Tools

read_file

Read the contents of a file in the workspace.

When it's useful: Reviewing code, reading configuration files, inspecting data files that were previously created or uploaded.

Example prompt:

Read the config.yaml file and explain what each setting does.

write_file

Create a new file or overwrite an existing file in the workspace.

When it's useful: Generating reports, saving code, creating configuration files, exporting analysis results.

Example prompt:

Create a requirements.txt with pandas, numpy, and scikit-learn.

list_files

List files and directories at a given path in the workspace.

When it's useful: Exploring the workspace filesystem, understanding project structure, finding files before reading them.

Example prompt:

What files are in the /data directory?

find_file

Search for files by name pattern across the workspace.

When it's useful: Locating files when you know part of the name but not the exact path.

Example prompt:

Find all CSV files in the workspace.

Git Tools

git_clone

Clone a Git repository into the workspace.

When it's useful: Pulling in codebases for review, importing starter templates, working with version-controlled projects.

Example prompt:

Clone the repo at https://github.com/example/analytics-pipeline into the workspace.

git_commit

Stage and commit changes to a Git repository in the workspace.

When it's useful: Saving progress after code edits, creating checkpoints in a cloned repo.

Example prompt:

Commit all changes with the message "Add data cleaning script".

Data Tools

Data tools require an active data connection of the matching type. If no connection is configured, the tool will return an error explaining what's needed.

query_bigquery

Run a SQL query against a connected BigQuery dataset and return the results.

When it's useful: Exploring data, running ad-hoc analytics, generating reports from your data warehouse.

Example prompt:

Show me the top 10 customers by revenue from the orders table this quarter.

query_snowflake

Run a SQL query against a connected Snowflake database and return the results.

When it's useful: Querying Snowflake data warehouses, joining across schemas, running analytical SQL.

Example prompt:

What's the average order value by region from the sales.public.orders table?

query_databricks

Run a SQL query against a connected Databricks workspace and return the results.

When it's useful: Querying Databricks lakehouses, exploring Unity Catalog tables, running Spark SQL.

Example prompt:

List all tables in the analytics catalog and describe the schema of the events table.

System Tools

System tools are always enabled and cannot be turned off. They provide workspace awareness and inter-workspace communication.

describe_workspace

Returns metadata about the current workspace, including its name, members, enabled tools, active connections, and configuration.

When it's useful: The AI uses this tool automatically to understand the workspace context. You can also ask about it directly.

Example prompt:

What tools are enabled in this workspace and who are the members?

bridge_workspace

Send a message to or delegate a task to another connected workspace. Bridges must be configured between workspaces before this tool can be used.

When it's useful: Multi-workspace workflows — for example, asking a data workspace to run a query and return results to a reporting workspace.

Example prompt:

Ask the data-pipeline workspace to run the daily ETL summary and send me the results.
info

Bridges are available on the Business plan and above. See Bridges for setup details.


Enabling and Disabling Tools

You can control which tools are available to the AI on a per-workspace basis. This is useful for restricting the AI's capabilities in sensitive environments — for example, disabling shell_exec in a workspace used by external stakeholders.

To manage tools:

  1. Open your workspace.
  2. Click the Settings icon (gear) in the workspace header.
  3. Navigate to the Tools section.
  4. Toggle individual tools on or off.
  5. Click Save.
tip

When you disable a tool, the AI will no longer attempt to use it. If a user asks the AI to do something that requires a disabled tool, the AI will explain that the capability is not available in this workspace.

Keep in mind:

  • System tools (describe_workspace and bridge_workspace) are always enabled and cannot be toggled off.
  • Data tools (query_bigquery, query_snowflake, query_databricks) require a matching data connection to function, even when enabled.
  • Tool changes take effect immediately for all new messages in the workspace.
  • Only workspace owners and admins can enable or disable tools.