Support
Log In

Connect your Coding Agent

Let any coding agent analyze data with your published models

The modeling MCP tools are for building models. To let your coding agent analyze data with your published models — the same governed access as workspace chat — connect it to the consumption MCP server and authenticate with OAuth.

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. To narrow an agent to one organization or one workspace, open Connect AI in the Credible app (settings gear in the sidebar, then Connect AI), pick the narrower scope under Access, and follow the steps it shows — they carry the scoped URL (see MCP Tools for the endpoint shapes). Every client below takes the same URL — only the syntax changes.

Building models rather than querying them? The VS Code Extension writes the modeling MCP config and skills for you in VS Code, Cursor, and Claude Code — no URL to paste. Connecting a personal chat client like Claude, ChatGPT, or Gemini? See Connect your Agent.

The Credible plugin installs the connection and the analysis skills that teach Claude how to use it. These are slash commands — type them inside a Claude Code session:

/plugin marketplace add anthropics/claude-plugins-community
/plugin install credible@claude-community

Then run /mcp and sign in to Credible.

Anthropic's community marketplace is not pre-added — only the official one is — so the first command is required, not optional.

For the connection without the skills — this one is a shell command, run in your terminal rather than inside a session:

claude mcp add --transport http credible https://mcp.credibledata.com/global/

Run these in your terminal before you start Codex. They write to Codex's config, and a session that is already running won't pick the server up until it restarts:

codex mcp add credible --url https://mcp.credibledata.com/global/
codex mcp login credible

Both write to ~/.codex/config.toml, which you can edit directly instead:

[mcp_servers.credible]
url = "https://mcp.credibledata.com/global/"
auth = "oauth"

Remote servers are first-class in current Codex. Older builds read only stdio servers and skip this one without saying so — if nothing shows up, upgrade Codex, or add this above the server entry:

[features]
experimental_use_rmcp_client = true

In the Codex IDE extension or app there is no terminal step: Settings → MCP servers → Add server → Streamable HTTP, paste the URL, save, then Authenticate.

  1. Press Cmd+Shift+P (Ctrl+Shift+P on Windows/Linux) and run "Open MCP Settings"
  2. Click Add MCP Server
  3. Add the configuration:
{
  "mcpServers": {
    "credible": {
      "url": "https://mcp.credibledata.com/global/"
    }
  }
}
  1. Save — Cursor opens the OAuth flow; complete it in your browser

Cursor marks a server as remote by the presence of url, so there is no type key here. Put the config in .cursor/mcp.json for one project, or ~/.cursor/mcp.json for every project.

  1. Open the Command Palette (Cmd+Shift+P on Mac, Ctrl+Shift+P on Windows/Linux) and run "MCP: Add Server"
  2. Select HTTP
  3. Enter the MCP URL: https://mcp.credibledata.com/global/
  4. Enter a server ID (e.g., "credible")
  5. Press Enter and select "Global" or "Workspace" scoped
  6. A popup redirects you to the OAuth flow — open it in your browser and complete the authentication
  7. The server now shows as "Running" (with a checkbox) in the mcp.json file
  8. Open Copilot to test it out

To write the config by hand, use .vscode/mcp.json for one workspace, or run "MCP: Open User Configuration" for every workspace:

{
  "servers": {
    "credible": {
      "type": "http",
      "url": "https://mcp.credibledata.com/global/"
    }
  }
}

Add the server to opencode.json:

{
  "$schema": "https://opencode.ai/config.json",
  "mcp": {
    "credible": {
      "type": "remote",
      "url": "https://mcp.credibledata.com/global/",
      "enabled": true
    }
  }
}

opencode starts the sign-in flow itself the first time the server answers with a 401, and keeps the tokens for later sessions.

Install the CLI, add the server, then start gemini — it opens a browser window for sign-in:

npm install -g @google/gemini-cli
gemini mcp add -t http credible https://mcp.credibledata.com/global/
gemini

gemini mcp list confirms the server is connected.

  1. In the agent window, click on the plug icon
  2. Click Settings — this opens ~/.codeium/windsurf/mcp_config.json, which you can also create and edit directly
  3. Add your server configuration:
{
  "mcpServers": {
    "credible": {
      "serverUrl": "https://mcp.credibledata.com/global/"
    }
  }
}
  1. Save the file — this automatically opens the OAuth authentication flow
  2. Complete the authentication in your browser

Any other MCP client works the same way: Credible is a remote MCP server over Streamable HTTP, and sign-in is OAuth with dynamic client registration — the client registers itself, so there is no key to copy and no client ID to configure.

Once connected, your agent has get_context and execute_query — the same tools that power workspace chat — plus list_workspaces, search_malloy_docs, and get_skill. See the MCP Tools reference for what each one takes.

Next Steps

On this page