Quickstart
Five steps from a new Credible organization to a published data model you can query from the Credible App, your favorite agent and anything you build
When you first sign in, the Credible agent walks you through these steps in a chat: connect a database, upload a file or start with sample data, then build and publish your first data model. This page lists the same steps, so you know what the agent is doing, or can do them yourself.
Set Up in the Credible App
Create an organization
Pick one:
-
In your browser: sign up (opens in a new tab) with Google, GitHub or Microsoft.
-
In a terminal or coding agent:
npm install -g @credibledata/cred-cli && cred signup -
From Claude: add the Credible connector (opens in a new tab) and ask a question. Credible replies with a signup link.
If your company already uses Credible, ask an admin to invite you. More on signing up.
Connect your data
Your data lives in an environment, which holds your connections, your data models and who can use them. If you don't have one yet, the agent creates one when you connect.
Add a database or warehouse, point Credible at files in cloud storage, or upload a file. Only environment admins can add connections. If you have no data to hand, the agent's connection form offers a sample dataset. See Connect Data.
Build a data model with the agent
Tell the agent what you want to answer. It explores your tables, proposes sources, joins and measures, and writes a documented Malloy model in a draft, compiling each file as it goes. Ask it to run a few queries so you can check the numbers before you publish. See Build a Model, and The Data Model for what goes into a model.
Publish it
Ask the agent to publish the draft. Publishing needs the Admin or Modeler role on the environment and a model that compiles. It creates a package with its first version, and that version is what people, agents and apps read by default. See Publishing and Versions.
Query and share the model
Connect Claude, ChatGPT and other chat clients or a coding agent to Credible over MCP, and ask questions there. In the app, chat with your data, save answers as reports, and build data apps. To let others in, share a workspace or grant roles on the environment. See Analyze and Share.
Set Up with a Coding Agent
Paste this prompt into Claude Code, Codex, Cursor or another coding agent.
Set up Credible following https://www.credibledata.com/docs/how-to/developers/local-modeling and build me a first data model of my data.Next Steps
Row and Column Access
Limit who sees which rows, columns and packages
Materialization and Performance
Speed up queries and lower warehouse cost
Descriptions and Value Search
Describe fields so people and agents pick the right ones
Local Development
Work in your own editor and coding agent, with Git
Migrate to Credible
Bring definitions over from Looker, Tableau, Power BI and others
How Credible Works
The ideas behind these steps
Have questions? Get in touch: hello@credibledata.com