Get Started

What Credible is, who these docs are for, and a map of the documentation sections with a link to each

Credible is the AI Analytics Engine. You connect your data, write a data model in Malloy that says what the data means, and Credible builds and serves that model to people, apps and agents.

What Credible Does

The data model is the input you write and own: sources, joins, dimensions, measures, documentation and access rules, in version-controlled Malloy files. Credible is the engine that runs it. From a published model, Credible:

  • compiles each question into SQL and runs it against your data, live or from tables it materializes for you,
  • builds a search index over the model so people and agents find the right fields,
  • checks permissions and access rules on every query at one gateway, and
  • serves results to chats in the app, reports, data apps, Slack, AI agents over MCP, and the REST APIs.

Your data stays where it is: a database or warehouse, files in cloud storage, or files you upload. For how the parts fit together, read How Credible Works.

Who These Docs Are For

  • People who build data models: data and analytics engineers, analysts, and anyone working with the modeling agent in the app or with a coding agent on their own machine.
  • Administrators: people who set up connections, environments, users and access.
  • Developers: people who build data apps, agents and integrations on a published model through MCP and the APIs.
  • People who analyze data: anyone asking questions in a workspace, in Slack or from an AI assistant.

Credible comes as Open Source (you run Malloy Publisher (opens in a new tab), the open-source engine Credible maintains, on your own machines), Cloud (managed by Credible) and Enterprise. These docs describe the managed product. See Pricing for what each plan includes.

One Set of Skills, Every Surface

Credible's agents follow agent skills: readable instructions for exploring data, writing Malloy, analyzing without guessing, building data apps and publishing. The Malloy skills, and the MCP tools they call, are open source in Malloy Publisher (opens in a new tab); the Credible-specific skills, for things like publishing to Credible and the app's onboarding, are not. The in-app agent, the local setup cred init creates for a coding agent, the VS Code extension and the Credible plugin all draw on the same skills, so an agent works the same way on each surface. You can read the open-source ones, fork them, or extend them with your team's own knowledge.

How the Docs Are Organized

The sidebar follows the order you use Credible in: connect data, model it, control access, then deliver it.

SectionWhat it covers
Get StartedThe Quickstart, connecting AI agents (Claude, ChatGPT, coding agents) over MCP, and the key concepts (how Credible works, the data model, environments and packages, workspaces, and Malloy).
Connect DataConnecting databases and warehouses, files in cloud storage, and uploaded files.
Model Your DataBuilding a model with the agent or migrating one from another tool, documenting it, row and column access, materialization, publishing and versions, and git-backed environments.
Control AccessUsers and groups, roles and permissions, and single sign-on.
Analyze and ShareChatting with your data, reports, data apps, shared workspaces and Slack.
Local DevelopmentModeling in your own editor with the cred CLI and the VS Code extension, and CI/CD.
AdministrationDeployment options, monitoring, and setting up sign-in with Microsoft Entra ID.
ReferenceThe CLI, the REST APIs, the annotations, the glossary and Get Help.

Questions? Contact support@credibledata.com.

Next Steps

On this page