Developer tools: the connector, the CLI and the SDK

What the Data Visions connector, command line and Module SDK are, who uses which, how they fit together, and where to start.

Copy page for AIUpdated November 21, 2026

Data Visions gives you three tools for working with your workspace from outside the app: the connector, the command line (CLI) and the Module SDK. They all act as you, in one workspace at a time, with exactly what your role allows. This page explains what each one is, who uses which, and what you need before you start.

The three tools

ToolWhat it isRuns where
The connectorA connection (an MCP server) that you add to an AI app. The app can then look things up, run reports, change records and, for Workspace Admins, build and deploy modules.Inside Claude (claude.ai, Claude Desktop, Claude Code), ChatGPT, Cursor and Codex
The command line (CLI)The dv program on your computer. It does the same module work as the connector from a terminal, and a little more.A terminal on Windows, macOS or Linux
The Module SDKThe guides, skills, references and an example module that teach a coding agent how to build a module the way the platform expects.Read through the connector's sdk_doc tool

Who uses which

You areUseWhat it looks like
A person in Claude or ChatGPT who wants answers from your dataThe connector"Show my open work orders." "Make a report of last month's hours by project."
A Workspace Admin who wants a module built or changed, without writing codeThe connector, in Claude or ChatGPT"Check out our Tool Calibration module and add a due-date reminder."
A coding agent such as Claude Code, Cursor or Codex, building a module in a folder on your computerThe connector for the SDK, checks and deploys, plus the CLI if the module is largeThe agent reads the SDK, writes the files, runs the module check and asks you before it deploys.
A developer at a terminalThe CLIdv modules checkout, dv modules review, dv modules deploy

Most people never need the CLI. If you only want to ask questions and change records, the connector is all you need.

How they fit together

  • The connector serves the SDK. There is no SDK download to keep up to date. A Workspace Admin's AI app reads the current SDK through the sdk_doc tool, starting with skills/starting-a-module/SKILL.md. The SDK always matches the platform you are on.
  • The connector and the CLI do the same module work. Both can check a module out and in, run the module check, deploy it to your workspace, show its versions and submit it to the Marketplace. Use whichever suits where you are working, and mix them: a check-in from the CLI and a check-in from Claude both go into the same module draft.
  • The CLI handles what the connector can't. The connector takes a module's files as part of a message, so it accepts up to 400 files and about 8 MB (2 MB a file). Larger modules are deployed with the CLI, which uploads from a folder on your computer.
  • The connector can set up the CLI. Ask your AI app to set up the Data Visions CLI. Its cli_setup tool gives per-user install steps for your computer (no administrator rights needed) and a download link made for you that works for 15 minutes.
  • All three meet the same rules. Nothing goes live until it passes the module check, and a deploy only ever goes into the workspace you are signed in to.

What you need first

  1. An account in the workspace. Each person signs in as themselves. No one shares a login or a key.
  2. The workspace's AI connectors setting. A Workspace Admin chooses Off, Read only or Read and make changes under System settings › Security › AI connectors. Off stops every connected app. Read only stops apps from creating or changing records. See Security: AI connectors.
  3. The Workspace Admin role, for building modules. Reading the SDK, checking, deploying, checking out and checking in modules are for Workspace Admins only, through the connector and the CLI alike. Everyone else can use the connector for their own data.
  4. The connector address. A Workspace Admin finds it under System settings › Marketplace › Connect your AI, and clicks Copy. Share it with anyone who wants to connect.

Open the Marketplace now

Where to start

I want toRead
Connect my AI app and understand what it can doThe Data Visions connector (MCP)
See every tool the connector offersConnector tool reference
Install the CLI and sign inInstall and sign in to the CLI
Look up a CLI commandCLI command reference
Know what the SDK contains and how an agent reads itThe Module SDK: guides and references for coding agents
Build a module with Claude Code, Cursor or CodexBuild a module with a coding agent, start to finish
Look up a field in module.jsonModule manifest (module.json) reference

New to building? Start with Building on Data Visions: an overview. The Module Builder inside the app needs none of these tools.

Good to know

  • None of these tools is more powerful than you. Each one acts as the person who signed in, with that person's role.
  • Connecting an AI app doesn't need a token. You sign in on the Data Visions page and choose Allow. The CLI signs in with your password, typed by you in the terminal.
  • Several workspaces? The connector and the CLI work in one workspace at a time. Check which one before you change anything.
  • Modules from the Marketplace can't be checked out. Their code belongs to their publisher. You can check out only modules your workspace owns.
  • Everything is recorded. Every connector tool call is written to your workspace's audit log.

Can't find what you need? Contact support.