Start here

Get started with SignalPilot

SignalPilot sits between your AI tools and your data. Ask a question in Claude, ChatGPT, or SignalPilot's own chat, and it runs a checked, read-only query against your warehouse and shows its work.

Open SignalPilotHow it works

1. 1

   Create your workspace

   Sign up at app.signalpilot.ai. If a colleague invited you, accept the invitation and you land in their workspace with its connections and definitions already in place. Otherwise, name your workspace after your team. The free plan is enough to try everything on this page.

2. 2

   Connect a database

   Open Connections → Add connection, choose your warehouse, and paste read-only credentials. SignalPilot supports Snowflake, BigQuery, Postgres, Redshift, Databricks, SQL Server, and more. The connection is tested before it is saved.

   No database handy? Turn on the demo warehouse and try everything against sample data first.
3. 3

   Ask a question you can check

   Open Chats, pick the project, and ask something you already know the answer to, such as last month's order count. Read the answer, then open the evidence dropdown to see the SQL. If the number is right, you are set. If not, go to the next step.
4. 4

   Teach it your definitions

   Open the knowledge base and write down the things your team knows but the schema does not: what counts as an active customer, which table is the source of truth for revenue, which fiscal calendar you use. Every later answer uses them automatically.
5. 5

   Add SignalPilot to the AI tool you already use

   You can stay in SignalPilot's chat, or bring the same governed access into your everyday tools.

   ClaudeAdd SignalPilot as a connector in claude.ai.ChatGPT and OpenAIChatGPT connectors, Codex, and the Agents SDK.Claude CodeOne command, plus skills for dbt and SQL work.Cursor and other toolsAny tool that supports MCP.

Nothing is written to your database

SignalPilot only ever reads. Statements that would change data are blocked before they run, results are capped, and every query is recorded in the audit log.

Where to go next

ProductChat with your dataWhat happens during a run, and how answers are built.ProductKnowledge baseThe step that makes answers match your business.ExtendAdd MCP serversGive the chat agent tools from Linear, GitHub, Notion, and more.TrustGovernanceExactly what is blocked, limited, and logged.TeamsEvalsGrade the agent on questions you know the answer to.EngineersSelf-hostRun the whole stack yourself with Docker Compose.