Credible home
Support
Log In

Connect your Agent

Connect Claude, ChatGPT, Gemini, and other MCP-compatible chat clients to your data models

Ask the same questions of the same governed models you analyze in the Credible App, from the agent you prefer — Claude, ChatGPT, Gemini, or any MCP-compatible chat client. Connect it to Credible and it answers questions grounded in your governed data models — no hallucinated field names, no misread schemas — powered by the same AI Analytics Engine and the same open-source agent skills behind workspace chat. Same analysis discipline, your agent.

Personal agents connect to Credible's MCP server and authenticate with OAuth: add the server, complete the sign-in flow as yourself, and start asking questions. Because you sign in as yourself, every query is governed as you — your model's access rules apply exactly as they do in the app.

Use the Credible MCP server URL: https://mcp.credibledata.com/global/

One URL covers every organization and workspace your account can reach — the server resolves scope from who you signed in as, so there is nothing org-specific to look up. To scope an agent down to one organization or one workspace instead, open Connect AI in the Credible App and pick the narrower scope under Access; the setup steps it shows carry the scoped URL. See MCP Tools for the endpoint shapes.

Connect AI is the quickest route to everything on this page: it prints the setup steps for each client with your URL already filled in, and every role can reach it. Open it from the settings gear in the sidebar, then Connect AI; a workspace's settings page links straight in with that workspace already selected.

Using an IDE or CLI coding agent — Claude Code, Codex, Cursor, VS Code Copilot, opencode, Gemini CLI, or Windsurf? See Connect your Coding Agent for connecting it to your published models.

The Credible plugin installs the connection and the analysis skills that teach Claude how to use it — phrase detection, Malloy query patterns, chart selection, and the analysis workflow. It works here, in the Chat tab of Claude Desktop, and in Cowork:

  1. Open Customize in the sidebar, then the Plugins tab
  2. Under Personal plugins, click + and choose Add marketplace
  3. Enter credibledata/credible-plugin, then click Sync
  4. Install Credible from the list
  5. Open the installed plugin, go to its Connectors tab, and click Connect

Step 5 is easy to miss and is the reason a correct install can show no data: the plugin's connector is a second, separate install, and connecting it is what triggers sign-in.

Use credibledata/credible-plugin here, not the community catalog. Credible is listed in anthropics/claude-plugins-community as well, and that is the marketplace Claude Code adds — but the same slug fails in this dialog with "Failed to add marketplace". The plugin's own directory page is not a way around it: remove the marketplace and that page stops installing too.

Connector only, no skills. Credible is in Anthropic's connector directory, so there is nothing to paste:

  1. Open claude.ai/directory/credible, or find Credible under Customize → Connectors → Browse connectors
  2. Click Connect and sign in to Credible

The connector works across claude.ai, Claude Desktop, and the mobile apps — you add it once per account. A connector carries the tools; the plugin above carries the tools and the skills.

To add it by URL instead — a narrowed scope from Connect AI, or a deployment other than production:

  1. Open Customize → Connectors
  2. Click +, then Add custom connector
  3. Enter Name: Credible and the remote MCP server URL: https://mcp.credibledata.com/global/
  4. Leave the OAuth fields empty and click Add
  5. Click Connect on the next page and complete the sign-in flow

On Cloud and Enterprise plans, members cannot add connectors themselves. Only an Owner or the Primary Owner can, under Organization settings → Connectors (Add, then Custom → Web, then the URL). Each member then connects and signs in individually under Customize → Connectors — there is no org-wide pre-authentication. On a Cloud plan, the directory listing shows members a Request button that sends it to those Owners.

Cowork installs the same plugin as Claude chat, and gets the same connection and analysis skills. Follow the plugin steps on the Claude tab, with one difference: open the Cowork tab first, then Customize.

Cowork also shares your Claude account's connectors, so a connector added on claude.ai works here too.

Requires developer mode, available on Plus, Pro, Business, Enterprise, and Education accounts, on the web.
  1. On the web app, open Settings → Security and login and turn on Developer mode
  2. Open Plugins, click +, and create a developer-mode app for a remote MCP server
  3. Enter a name (e.g., "Credible") and a description
  4. Enter the MCP server URL: https://mcp.credibledata.com/global/
  5. Leave authentication as OAuth and create the connection
  6. Complete the OAuth sign-in flow
  7. Now you can ask ChatGPT to use the Credible tool (or whatever you named it)

On a Business, Enterprise, or Education workspace, an admin has to allow developer mode first, under Workspace settings → Permissions & roles → Connected data. On Business plans a published app cannot be edited afterwards — to change it, recreate and republish.

OpenAI has renamed this surface more than once — developer mode and custom MCP apps have lived under Apps, Apps & Connectors, and now Plugins. If a label here doesn't match what you see, look for the nearest equivalent rather than assuming the feature is gone.

The consumer Gemini web app does not support custom MCP connectors. Use Gemini Enterprise, or connect the Gemini CLI instead.

In Gemini Enterprise, connect Credible as a custom MCP server data store:

  1. In the Gemini Enterprise console, go to Data Stores → Create Data Store and select Custom MCP Server
  2. Enter the MCP Server URL: https://mcp.credibledata.com/global/, along with the OAuth configuration for your organization
  3. Attach the data store to your Gemini Enterprise app under Connected data sources
  4. In the Gemini chat input, open the connector menu, Authorize the Credible connector, and toggle it on

Gemini Enterprise requires OAuth client credentials to authorize the connection — contact support and we'll help you get set up.

Connecting your own agent over MCP is not metered: your agent, your tokens. Users, seats, and MCP access are free on every plan — see pricing.

What Your Agent Can Do

Once connected, your agent has the tools every Credible surface uses — get_context to discover governed entities in your models and execute_query to run Malloy queries against them, plus list_workspaces to see what it can reach, search_malloy_docs for language reference, and get_skill to read Credible's open-source skills on demand. Those skills encode the discipline to use the tools well: parsing your question into the right search phrases, finding the right sources and views, writing correct Malloy, choosing sensible visualizations, and sanity-checking results before trusting them. In your agent, that means you can:

  • Ask data questions in plain language — "What were our top products by revenue last quarter?" — and get answers grounded in your governed definitions, not guesses
  • Iterate in conversation — follow up, refine, and drill into drivers, with each query building on the matched entities from the last
  • Analyze alongside your other context — combine Credible answers with the documents, tools, and conversation already in your agent
  • Stay governed throughout — every query runs through Credible, ACL'd and audit-logged as you

If your agent has several workspaces to choose from, it will ask which one to work in the first time; tell it once and it remembers.

See the MCP Tools reference for tool parameters and details.

Build Data AppsConnect with Slack

On this page

What Your Agent Can Do