> ## Documentation Index > Fetch the complete documentation index at: https://docs.credibledata.com/llms.txt > Use this file to discover all available pages before exploring further. # Analyze Data > Chat with your data in governed workspaces in the Credible App The Credible App (`https://.app.credibledata.com`) is one way your team can consume published semantic models. In the Credible App, work happens inside a **workspace** — your personal workspace, or a shared workspace scoped to a set of packages and team members (see [Creating a Workspace](#creating-a-workspace) below). Workspace landing page showing Data Apps, Reports, Models, and Conversations ## Chat Type a question in the chat bar at the top of any workspace. The agent parses your question and uses the [Credible Analytics Engine](/how-to/analyzing/overview#how-it-works) to search your published semantic models — matching phrases to governed entities (dimensions, measures, views) and indexed dimension values via their `#(doc)` and `#(index)` annotations — then constructs and executes a trustworthy Malloy query against your data. get_context tool showing phrase parsing and entity matching Each chat is saved in the workspace and visible to all members — find previous conversations under **Conversations**. ### How It Works 1. **`get_context`** — The agent parses your question into semantic phrases and calls `get_context` to search your published semantic models and indexed dimension values, getting back matched data entities. Match quality depends on the `#(doc)` descriptions and `#(index)` annotations in your model. 2. **`execute_query`** — The agent constructs and runs a Malloy query using the matched entities, returning results and visualizations. This is what makes answers trustworthy — the agent is grounded in governed definitions, not guessing column names or interpreting ambiguous schemas. And how it analyzes — the query patterns and the rigor — comes from Credible's [open-source agent skills](/introduction#one-set-of-skills-every-surface): the same skills every Credible surface runs, curated in the open with the world's data experts. Published models must be [indexed](/how-to/modeling/publishing#what-happens-when-you-publish) before chat works. Check that the package version shows **Ready** on its package page. ## Creating a Workspace Click **+ New Workspace** in the left sidebar of the Credible App. The creation wizard has three steps: 1. **General** — Name your workspace and add an optional description. Choose a name that reflects the team or use case (e.g., "Ecommerce Analysis", "Marketing Metrics") 2. **Members** — Add users or groups, each with a role: **Manager** (full control over workspace settings, members, and packages) or **Member** (can view models, create chats, and create reports) 3. **Packages** — Add the published semantic model packages you want available in this workspace. Only packages shared with members (via [environment permissions](/platform-admin/permissions#sharing-environments--packages)) can be added Workspace managers can update settings at any time — add or remove members, change roles, add or remove packages, or delete the workspace. Navigate to your workspace and click **Settings**. ## See Also * [Build Data Apps](/how-to/analyzing/data-apps) — Build and use interactive dashboards and applications shipped with your packages * [Build & Publish](/how-to/modeling/in-app-development) — Build and publish a model with the agent, then chat with it * [Analytics Engine Overview](/how-to/analyzing/overview#how-it-works) — How phrase-matching and entity retrieval work * [Connect your Agent](/how-to/analyzing/connect-your-llm) — Chat with the same models from Claude, ChatGPT, Gemini, or any MCP client * [MCP Tools](/how-to/analyzing/ai-assistants-mcp) — Technical details on `get_context` and `execute_query` for custom agents * [Permissions](/platform-admin/permissions) — Sharing workspaces and documents, and the full permission model