Available Tools

Each Ballast collection exposes MCP tools that Claude can use. The available tools depend on your collection’s data sources.

Document Collection Tools

Collections with non-federated sources (Slack, Google Drive, Notion, etc.) expose these tools:

search

Search the collection with full parameter control.

Parameters:

Name	Type	Required	Description
query	string	Yes	Search query
limit	int	No	Max results (default: 10, max: 20)
offset	int	No	Pagination offset
search_type	string	No	semantic, keyword, or hybrid (default)
expand_query	bool	No	Generate query variations
rerank	bool	No	Apply LLM reranking
rerank_model	string	No	cohere or jina
generate_answer	bool	No	Synthesize answer from results
temporal_relevance	float	No	Recency weighting (0.0-1.0)
interpret_filters	bool	No	Extract filters from natural language
filters	object	No	Explicit metadata filters

Example:

{
  "query": "remote work policy",
  "limit": 10,
  "rerank": true
}

list_documents

List documents with metadata only (no content). Efficient for browsing.

Parameters:

Name	Type	Required	Description
limit	int	No	Max documents (default: 20, max: 100)
offset	int	No	Pagination offset

read_document

Read a specific document’s content with optional pagination.

Parameters:

Name	Type	Required	Description
id	string	Yes	Document UUID
offset	int	No	Character offset to start from
limit	int	No	Max characters to return

Use pagination for large documents to save tokens.

list_sources

List all source connections in the collection.

Returns: Array of source connections with status and last sync time.

trigger_sync

Trigger a data sync for a specific source connection.

Parameters:

Name	Type	Required	Description
source_connection_id	string	Yes	Source connection UUID

get_info

Get collection metadata including name, description, and settings.

Database Collection Tools

Collections with federated database sources expose additional tools:

list_tables

List all tables in the connected database.

Returns: Array of tables with schema and name.

describe_table

Get the schema for a specific table.

Parameters:

Name	Type	Required	Description
schema	string	No	Database schema (default: public)
table	string	Yes	Table name

Returns: Array of columns with name, data type, and nullability.

query_data

Execute a read-only SQL query.

Parameters:

Name	Type	Required	Description
query	string	Yes	SQL SELECT query
limit	int	No	Max rows (default: 100, max: 1000)

Only SELECT statements are allowed. The query runs directly against your database for live results.

Visualization Tools

All collections expose:

get_dashboard_conventions

Get industry-specific best practices for visualizations.

Parameters:

Name	Type	Required	Description
industry	string	No	Industry name (auto-detected if omitted)

Supported industries: finance, healthcare, ecommerce, marketing, saas, devops, manufacturing, hr

create_chart

Create and save a visualization chart.

Parameters:

Name	Type	Required	Description
chart_type	string	Yes	Chart type (line, bar, pie, etc.)
title	string	Yes	Chart title
data	object	Yes	Chart.js compatible data
source_tool	string	Yes	Tool used to get data (query_data or search)
source_args	object	Yes	Arguments passed to data tool (for refresh)
options	object	No	Chart.js options

Available chart types: line, bar, pie, doughnut, area, scatter, bubble, radar, polarArea, horizontalBar, stackedBar, stackedArea, gauge, heatmap, treemap, funnel

Resources

Collections also expose MCP resources:

ballast://collection

Collection details including ID, name, and settings.

ballast://sources

List of source connections in the collection.