Tools Reference

All 53 SignalPilot MCP tools across 12 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, 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.

Tool	Role in this stage
get_knowledge	Load baseline org/project knowledge before the scan
list_database_connections	Confirm the target connection exists by name
connection_health	Confirm the connection is reachable before querying
list_tables	Enumerate available tables in the connection
schema_overview	High-level shape: table count, rows, FK density, hub tables
schema_statistics	Aggregate table sizes and FK connectivity
get_relationships	FK/ERD map of the project
get_date_boundaries	Data-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.

Tool	Role in this stage
query_database	COUNT(*) / COUNT(DISTINCT key) for grain; SELECT DISTINCT on status/flag columns
explore_table	Deep-dive an upstream table's columns, FK refs, and samples
explore_columns	Inspect specific columns' types, stats, and samples
explore_column	Distinct values for a single categorical column (drives CASE WHEN vocabulary)
describe_table	Column types and nullability for an upstream
schema_link	Find tables relevant to the task question (NL → schema)
find_join_path	Derive multi-table join paths
analyze_grain	Confirm candidate-key uniqueness / fan-out on upstreams
compare_join_types	Preview INNER/LEFT/RIGHT/FULL row counts to choose the JOIN
query_history	Reuse prior successful query patterns
search_knowledge	Pull task-relevant prior research

Write

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

Tool	Role in this stage
generate_sql_skeleton	Scaffold a model's SQL shape from its YML column list
validate_sql	Syntax/semantic check against live schema before materializing
explain_query	Plan check before running
estimate_query_cost	Dry-run cost guardrail on expensive queries
check_budget	Remaining session query budget
query_database	Spot-check intermediate logic while building
debug_cte_query	Localize errors inside a WITH query during build
dbt_error_parser	Turn 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.

Tool	Role in this stage
check_model_schema	Actual vs YML columns (missing/extra/case)
validate_model_output	Post-build row count vs source/expected (fan-out, empty model)
audit_model_sources	Full cardinality + NULL/constant column + grain audit
verify_model_values	Model aggregates vs independent COUNT(*)/COUNT(DISTINCT) on sources
analyze_grain	Re-confirm output grain/uniqueness
compare_join_types	Diagnose a row-count FAIL traced to JOIN choice
query_database	Verifier subagents' read-only cross-checks
dbt_error_parser	Parse errors from rebuilds triggered by FAILs
propose_knowledge	Record 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.

Tool	Role
list_workspace_projects	List notebook workspace projects
run_notebook	Execute an analysis notebook in an isolated cloud sandbox
list_notion_integrations	List configured Notion integrations
notion_search	Search Notion source pages
notion_fetch_page	Fetch full content of a Notion page
notion_create_page	Publish a report page under the configured destination
connector_capabilities	Connector tier / feature-flag probe (any stage)

---

Full catalog by category

Query Intelligence (7 tools)

See Query Intelligence tools.

Tool	Description
query_database	Execute governed, read-only SQL with auto-LIMIT, DDL blocking, dangerous function blocking, audit, PII redaction
validate_sql	Syntax + semantic validation via EXPLAIN (no execution, no budget charge)
explain_query	Execution plan with row estimates and cost warnings
estimate_query_cost	Dry-run cost estimate before executing (BigQuery: exact bytes)
debug_cte_query	Break CTE query into steps, validate each independently
check_budget	Remaining query budget for a session
query_history	Recent successful queries for a connection (session memory)

Schema & Exploration (10 tools)

See Schema & Exploration tools.

Tool	Description
list_tables	Compact one-line-per-table overview: columns, PKs, FKs, row counts
describe_table	Full column details: types, nullability, PKs, annotations, PII flags
explore_table	Deep-dive: column details + FK refs + sample values + referenced tables
explore_column	Distinct values with counts, NULL stats, optional LIKE filter
explore_columns	Multi-column stats: distinct counts, uniqueness, min/max/avg, samples
schema_overview	Database-wide summary: table count, total rows, FK density, hub tables
schema_ddl	Full schema as CREATE TABLE DDL with FK constraints
schema_statistics	High-level stats: table sizes, FK connectivity (sorted by row count)
schema_diff	Compare current schema against last cached version (DDL changes)
get_date_boundaries	MIN/MAX dates across all DATE/TIMESTAMP columns

Relationships (3 tools)

See Schema & Exploration tools.

Tool	Description
get_relationships	Full ERD: all FK relationships as arrows or adjacency list
find_join_path	FK-based join path discovery between two tables (1–6 hops)
schema_link	Find tables relevant to a natural-language question (NL → schema)

dbt & Verification (8 tools)

See dbt tools.

Tool	Description
dbt_error_parser	Parse raw dbt error output into structured diagnosis + fix suggestion
generate_sql_skeleton	Generate a SQL template from YML column spec + ref tables
check_model_schema	Compare materialized columns vs YML-declared columns
validate_model_output	Row count validation + fan-out detection + empty model detection
verify_model_values	Cross-validate model aggregate values against raw source data
audit_model_sources	Upstream audit: source row counts, fan-out/over-filter ratios, NULL scan
analyze_grain	Cardinality analysis: per-key distinct counts, fan-out factors
compare_join_types	Compare row counts across INNER/LEFT/RIGHT/FULL OUTER JOIN

Workspaces (2 tools)

See Operational tools.

Tool	Description
list_workspace_projects	List the dbt/notebook projects in the user's workspace
run_notebook	Run a .py notebook in an isolated cloud sandbox, return output + view URL

Knowledge Base (5 tools)

See Operational tools.

Tool	Description
get_knowledge	Load baseline docs + task-relevant knowledge entries
search_knowledge	Agent-directed search across the knowledge base
read_knowledge	Read one knowledge entry in full by ID
propose_knowledge	Propose a new knowledge entry, or edit one in place with overwrite=true
archive_knowledge	Archive (soft-delete) a knowledge entry by ID

Reporting (1 tool)

Tool	Description
manage_report	Create or permanently delete a rendered HTML report

Dashboards are artifacts/<name>.dashboard.json files written in the chat sandbox and checked with the sandbox tools dashboard_sample_data and dashboard_screenshot.

Semantic layer (2 tools)

Tool	Description
list_semantic_metrics	List metrics defined in the project's dbt semantic layer
verify_metric_conformance	Check that a model's aggregates conform to the metric definitions

Branch workflow — Xata (6 tools)

Branch-per-developer Postgres on Xata: fork a branch, work on it, diff it against the parent before merging.

Tool	Description
xata_list_branches	List branches on the connection's Xata project
create_xata_branch	Fork a new branch from a parent branch
delete_xata_branch	Delete a branch
xata_branch_diff	Diff two branches at the Xata control-plane level
schema_diff_branches	Compare table and column structure across two branches
get_dbt_profile	Emit a profiles.yml block for a branch's credentials

Model mapping (2 tools)

Tool	Description
map_columns	Map upstream columns against a model's YML contract; labels each column MAPPED, UNMAPPED-INCLUDE, or UNMAPPED-EXCLUDE
analyze_project_db	Summarise the database a dbt project builds into

Notion Integration (4 tools)

See Operational tools.

Tool	Description
list_notion_integrations	List configured Notion integrations with search scope and report destination
notion_search	Search Notion pages visible to an integration's access token
notion_fetch_page	Fetch full content of a Notion page by ID
notion_create_page	Create a page under the configured report destination

Connections (3 tools)

See Operational tools.

Tool	Description
list_database_connections	List all configured database connections
connection_health	Latency percentiles (p50/p95/p99), error rates, status per connection
connector_capabilities	Connector 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