Welcome to Credible

Credible is the AI context engine. Without context, AI gives confident answers no one trusts. Credible captures what your data means — the definitions, business rules, and relationships buried in docs, dashboards, SQL, and your experts’ heads — as governed semantic data models, then delivers your data and its meaning as context to every surface: AI agents, dashboards, data apps, and whatever you build next. Models are built on Malloy, an open-source language designed for capturing and querying data’s meaning. Your models are code — readable, versioned in Git, portable, and free from vendor lock-in.

​

How It Works

Everything in Credible follows one path from raw data to trusted answers:

1. Connect and model. Link your databases through managed connections — credentials stay in the platform, never on anyone’s machine. Then build semantic models with AI agents, in your browser or in your IDE.
2. Govern and publish. Models live in environments as versioned packages, with access control defined in the model and enforced on every surface, and materialization to optimize performance and cost. Publishing makes a model available to every consumer at once.
3. Deliver everywhere. Chat with your data and build reports in workspaces, give AI agents governed access via MCP (the open standard agents use to connect to tools), ship data apps with your models, or build on the REST APIs. One model, every surface, consistent answers.

​

One Set of Skills, Every Surface

Credible’s agents are not a black box. Their behavior is encoded as agent skills — readable playbooks for discovering data, modeling in Malloy, analyzing without hallucinating, building data apps, and publishing — and the MCP tools those skills use. Both are open source in Malloy Publisher, the open-source server for Malloy models. The in-app agent, the Credible Extension in your IDE, and the agent plugins we publish all run the same skills — so the agentic experience is consistent across every surface, and an answer arrived at in the app follows the same discipline as one in your coding agent. And because the skills are open, they’re curated with the world’s data experts — a community whose expertise runs deeper than any one vendor’s bench. Every skill is readable: fork them, or extend them with your organization’s institutional knowledge. Read the full story in We Open Sourced the Thing Everyone Else Is Selling.

​

Get Started

You can build semantic models, analyze data, and build data apps entirely in your browser or in your own IDE. Both produce the same governed packages — start with whichever fits you and mix them freely. Both paths use managed connections, publish to the same environments, and produce models ready for every consumer.

​

Explore the Docs

The documentation follows the same path — foundation first, then building, then consuming:

• Environments — Start here. The rest of this Get Started section covers the governed foundation: what environments are, how to connect your databases, and how to build and publish in the app.
• Semantic Data Modeling — Building models with AI agents, then evolving them: metadata, fine-grained access control, performance and cost, and publishing.
• Context Engine — Everything downstream of publish: analyzing data in workspaces, building data apps, and connecting your agent and Slack.
• Developers — The same foundation — same models, same skills, same MCP tools — in your own toolset: the VS Code extension, the CLI, CI/CD, MCP tools for custom agents, and the REST APIs.
• Admins — Managing your organization: users and groups, permissions and access control, and best practices for organizing environments and packages.
• Concepts — The ideas behind the platform: the semantic layer, the architecture, and why Malloy.

​

Next Step

Everything you build lives in an environment, so that’s the place to start:

---

Have questions or need assistance? Feel free to contact us at support@credibledata.com – we’re here to help!