Skip to main content

Tools Reference

All 40 SignalPilot MCP tools across 9 categories. The catalog below groups every tool by category (click a category for full parameter and example documentation). The Tools by workflow stage section then maps the same tools onto the four stages of a governed dbt/SQL run — Scan/Map, Explore/Research, Write, and Verify.

Tools by workflow stage

The dbt-workflow skill's steps collapse into four stages. Some scan/map and value-check operations run as bundled plugin CLI scripts (scan_project.py, validate_project.py, map-columns, verify-values) rather than MCP tools; the tables below list the MCP tools used at each stage by exact name. A tool can appear in more than one stage.

Scan / Map

Establish the shape of the database and project before writing anything: connections, table inventory, FK map, and data-driven date boundaries.

ToolRole in this stage
get_knowledgeLoad baseline org/project knowledge before the scan
list_database_connectionsConfirm the target connection exists by name
connection_healthConfirm the connection is reachable before querying
list_tablesEnumerate available tables in the connection
schema_overviewHigh-level shape: table count, rows, FK density, hub tables
schema_statisticsAggregate table sizes and FK connectivity
get_relationshipsFK/ERD map of the project
get_date_boundariesData-driven date floor/ceiling (replaces current_date hazards)

Plugin CLI in this stage: scan_project.py, validate_project.py.

Explore / Research

Confirm grain, discover categorical vocabularies, derive join paths, and preview JOIN behavior before committing to SQL.

ToolRole in this stage
query_databaseCOUNT(*) / COUNT(DISTINCT key) for grain; SELECT DISTINCT on status/flag columns
explore_tableDeep-dive an upstream table's columns, FK refs, and samples
explore_columnsInspect specific columns' types, stats, and samples
explore_columnDistinct values for a single categorical column (drives CASE WHEN vocabulary)
describe_tableColumn types and nullability for an upstream
schema_linkFind tables relevant to the task question (NL → schema)
find_join_pathDerive multi-table join paths
analyze_grainConfirm candidate-key uniqueness / fan-out on upstreams
compare_join_typesPreview INNER/LEFT/RIGHT/FULL row counts to choose the JOIN
query_historyReuse prior successful query patterns
search_knowledgePull task-relevant prior research

Write

Scaffold the model, validate and cost-check SQL against the live schema, and parse build errors while iterating.

ToolRole in this stage
generate_sql_skeletonScaffold a model's SQL shape from its YML column list
validate_sqlSyntax/semantic check against live schema before materializing
explain_queryPlan check before running
estimate_query_costDry-run cost guardrail on expensive queries
check_budgetRemaining session query budget
query_databaseSpot-check intermediate logic while building
debug_cte_queryLocalize errors inside a WITH query during build
dbt_error_parserTurn dbt run/parse stderr into a structured fix

Plugin CLI in this stage: map-columns, then dbt run --select <models>.

Verify

Post-build checks against YML contract and source data: column match, row counts, aggregate values, grain, and JOIN diagnosis.

ToolRole in this stage
check_model_schemaActual vs YML columns (missing/extra/case)
validate_model_outputPost-build row count vs source/expected (fan-out, empty model)
audit_model_sourcesFull cardinality + NULL/constant column + grain audit
verify_model_valuesModel aggregates vs independent COUNT(*)/COUNT(DISTINCT) on sources
analyze_grainRe-confirm output grain/uniqueness
compare_join_typesDiagnose a row-count FAIL traced to JOIN choice
query_databaseVerifier subagents' read-only cross-checks
dbt_error_parserParse errors from rebuilds triggered by FAILs
propose_knowledgeRecord learnings after the run

Plugin CLI in this stage: verify-values. Subagents: verifier, value-verifier.

Reporting / Notebooks (outside the four stages)

Used when the task asks for a notebook deliverable or a published report. connector_capabilities is a feature-flag probe usable at any stage.

ToolRole
list_workspace_projectsList notebook workspace projects
run_notebookExecute an analysis notebook in a sandboxed cloud pod
list_notion_integrationsList configured Notion integrations
notion_searchSearch Notion source pages
notion_fetch_pageFetch full content of a Notion page
notion_create_pagePublish a report page under the configured destination
connector_capabilitiesConnector tier / feature-flag probe (any stage)

Full catalog by category

Query Intelligence (7 tools)

See Query Intelligence tools.

ToolDescription
query_databaseExecute governed, read-only SQL with auto-LIMIT, DDL blocking, dangerous function blocking, audit, PII redaction
validate_sqlSyntax + semantic validation via EXPLAIN (no execution, no budget charge)
explain_queryExecution plan with row estimates and cost warnings
estimate_query_costDry-run cost estimate before executing (BigQuery: exact bytes)
debug_cte_queryBreak CTE query into steps, validate each independently
check_budgetRemaining query budget for a session
query_historyRecent successful queries for a connection (session memory)

Schema & Exploration (10 tools)

See Schema & Exploration tools.

ToolDescription
list_tablesCompact one-line-per-table overview: columns, PKs, FKs, row counts
describe_tableFull column details: types, nullability, PKs, annotations, PII flags
explore_tableDeep-dive: column details + FK refs + sample values + referenced tables
explore_columnDistinct values with counts, NULL stats, optional LIKE filter
explore_columnsMulti-column stats: distinct counts, uniqueness, min/max/avg, samples
schema_overviewDatabase-wide summary: table count, total rows, FK density, hub tables
schema_ddlFull schema as CREATE TABLE DDL with FK constraints
schema_statisticsHigh-level stats: table sizes, FK connectivity (sorted by row count)
schema_diffCompare current schema against last cached version (DDL changes)
get_date_boundariesMIN/MAX dates across all DATE/TIMESTAMP columns

Relationships (3 tools)

See Schema & Exploration tools.

ToolDescription
get_relationshipsFull ERD: all FK relationships as arrows or adjacency list
find_join_pathFK-based join path discovery between two tables (1–6 hops)
schema_linkFind tables relevant to a natural-language question (NL → schema)

dbt & Verification (8 tools)

See dbt tools.

ToolDescription
dbt_error_parserParse raw dbt error output into structured diagnosis + fix suggestion
generate_sql_skeletonGenerate a SQL template from YML column spec + ref tables
check_model_schemaCompare materialized columns vs YML-declared columns
validate_model_outputRow count validation + fan-out detection + empty model detection
verify_model_valuesCross-validate model aggregate values against raw source data
audit_model_sourcesUpstream audit: source row counts, fan-out/over-filter ratios, NULL scan
analyze_grainCardinality analysis: per-key distinct counts, fan-out factors
compare_join_typesCompare row counts across INNER/LEFT/RIGHT/FULL OUTER JOIN

Workspaces (2 tools)

See Operational tools.

ToolDescription
list_workspace_projectsList the dbt/notebook projects in the user's workspace
run_notebookRun a .py notebook in a sandboxed cloud pod, return output + view URL

Knowledge Base (3 tools)

See Operational tools.

ToolDescription
get_knowledgeLoad baseline docs + task-relevant knowledge entries
search_knowledgeAgent-directed search across the knowledge base
propose_knowledgePropose a new knowledge entry after a run

Notion Integration (4 tools)

See Operational tools.

ToolDescription
list_notion_integrationsList configured Notion integrations with search scope and report destination
notion_searchSearch Notion pages visible to an integration's access token
notion_fetch_pageFetch full content of a Notion page by ID
notion_create_pageCreate a page under the configured report destination

Connections (3 tools)

See Operational tools.

ToolDescription
list_database_connectionsList all configured database connections
connection_healthLatency percentiles (p50/p95/p99), error rates, status per connection
connector_capabilitiesConnector tier classification (Tier 1/2/3) and available features

Removed tools

The following tools were removed because the gateway runs in Docker and cannot access local filesystems. Their functionality is available via standalone scripts in the plugin:

  • execute_code / sandbox_status — sandbox disabled in cloud
  • dbt_project_map / dbt_project_validate — use scan_project.py and validate_project.py in the plugin
  • fix_date_spine_hazards / fix_nondeterminism_hazards — use the dbt-date-spines skill
  • create_project — use the web UI or API