signalpilot-dbt

SignalPilot Plugin for Claude Code

Governed AI database access — sandboxed queries, schema discovery, and intelligent model building powered by SignalPilot.

Install

Step 1: Connect the MCP server

# Cloud
claude mcp add --transport http signalpilot https://gateway.signalpilot.ai/mcp \
  --header "Authorization: Bearer sp_YOUR_API_KEY"

# Local / self-hosted
claude mcp add --transport http signalpilot http://localhost:3300/mcp

Step 2: Install the plugin (optional — adds skills + agents)

claude plugin marketplace add SignalPilot-Labs/signalpilot-plugin
claude plugin install signalpilot-dbt@signalpilot

Step 1 gives you all 30+ MCP tools. Step 2 adds skills and agents on top.

What's Included

MCP Tools (from Step 1)

Category	Tools
Schema Discovery	schema_overview, list_tables, describe_table, explore_table, explore_column, explore_columns, get_relationships, schema_ddl, schema_link
Querying	query_database, validate_sql, explain_query, estimate_query_cost
Analysis	analyze_grain, schema_statistics, find_join_path, compare_join_types, get_date_boundaries, schema_diff
Governance	check_budget, query_history, audit_model_sources, validate_model_output
Infrastructure	list_database_connections, connection_health, connector_capabilities
Notion	notion_search, notion_fetch_page, notion_create_page

Skills (from Step 2)

Skill	Description
/signalpilot-dbt:signalpilot	Main entry point — schema discovery, governed queries
/signalpilot-dbt:sql-workflow	Structured SQL query building with verification
/signalpilot-dbt:dbt-workflow	Full dbt project workflow (scan, map, validate, write, verify)
/signalpilot-dbt:dbt-write	dbt model writing with column naming and type rules
/signalpilot-dbt:dbt-debugging	Fix dbt run/parse failures
/signalpilot-dbt:dbt-date-spines	Fix current_date hazards in date spine models
DB-specific	bigquery-sql, snowflake-sql, duckdb-sql, sqlite-sql
/signalpilot-dbt:notion-context	Gather business context from Notion for dbt builds

Agents (from Step 2)

Agent	Description
verifier	Post-build verification of dbt models (7-check protocol)
notion-verify	Post-build traceability report — writes to Notion documenting how context influenced the build

How It Works

1. You ask Claude to build a dbt project or write SQL
2. Claude loads the signalpilot skill (tools overview + skill router)
3. For dbt projects, dbt-workflow orchestrates the workflow using SignalPilot MCP tools
4. If a Notion integration is configured in SignalPilot, business context is gathered before building (Step 0)
5. At Step 5, the verifier agent checks all models for correctness
6. If Notion context was used, a traceability report is written to Notion (Step 6)
7. You get a verified, working dbt project with full audit trail

Notion Integration (Optional)

Connect your Notion workspace to give the agent access to meeting notes, product specs, and data dictionaries for business context during dbt builds.

Setup

1. Open the SignalPilot web UI at http://localhost:3200/integrations
2. Click Add under the Notion section
3. Paste your Notion internal integration access token
4. Add page URLs for search scope (pages the agent can search)
5. Add a report destination URL (where verification reports are created)

Create a Notion integration at notion.so/profile/integrations to get an access token. Share pages with the integration in Notion for the agent to see them.

What it does

• Before build: Searches Notion pages for definitions, decisions, and constraints relevant to the task.
• During build: Agent references Notion context when making grain, join, filter, and column decisions. Leaves -- NOTION: <source> comments in SQL.
• After build: Writes a traceability report to Notion mapping every context item to the SQL decision it influenced.

Requirements

• Claude Code CLI
• A SignalPilot account (signalpilot.ai) or self-hosted instance

License

Apache-2.0