Meta note: this exact post was published using the datavessel CLI — in combination with Claude. Claude wrote the draft and drove the CLI to search for a royalty-free featured image, upload it to the media library, create the post, and publish it — all through datavessel’s own typed tools, straight from the terminal. Dogfooding, end to end.
From day one, datavessel has had a single idea behind it: you pilot, and your agents do the work. Connect your stack — Shopify, Shopware or WooCommerce, plus GA4, Search Console and Slack — and let agents analyze your data, run the work through your tools, and track what changed. You approve what matters; the rest runs on autopilot when you trust it.
That idea doesn’t belong to a single screen. The work happens in a chat window, in a script, on your phone, in your IDE. So we’ve been busy making datavessel meet you wherever the work actually is. Today we’re introducing the newest of those surfaces — the datavessel CLI — and sharing where the rest stand.
Introducing the datavessel CLI
The datavessel CLI brings your entire stack to the terminal. Install it in one line:
npm i -g datavessel-cliSign in once with your browser, and you have 100+ typed tools — across GA4, Search Console, Google & Meta Ads, Shopify, WooCommerce and Shopware — available as shell commands. It’s open source, scriptable, and built to be driven by both people and agents.
A few things make it more than a wrapper:
- Self-describing. Commands and flags are generated from the live datavessel tool catalog. Ship a new tool on the backend and it shows up in the CLI on the next sync — no release to install.
- Scriptable and pipeable. Every command takes
--json, so datavessel drops straight into cron jobs, CI pipelines and shell scripts:datavessel --json run … | jq. - Reads run free, writes ask first. Anything that touches the outside world prompts for confirmation unless you explicitly pass
--yes— the same approval model you get everywhere in datavessel. - Agent-ready. It ships a
SKILL.mdso coding agents like Claude Code and Cursor can install it and drive it directly. One file turns datavessel into a skill.
One backend, many surfaces
The CLI joins a growing family of ways to use datavessel — and they all sit on the same backend, the same typed tools, the same bring-your-own-key model. Different clients have different needs, so we meet each one on its own terms:
- MCP server — for conversational, exploratory work. Connect Claude Desktop or Mobile (or any MCP client) to
mcp.datavessel.io/mcpand ask questions in plain language; the model decides which tools to call and reasons across your whole stack. - CLI — for deterministic, scriptable, automation- and agent-driven work. When you (or a script, or an agent) know exactly what to run, the terminal is the fastest, most reproducible path.
- Web app — the home base, where you connect sources, manage agents and review what ran.
MCP and the CLI aren’t competitors — they’re two answers to the same question. Let the model use my stack, conversationally versus let me run my stack, deterministically. Most teams end up using both.
Datavessel in your pocket
Surfaces aren’t only about how technical you are — they’re about where you are. Agents keep working in the background, and the decisions that need you shouldn’t wait until you’re back at a desk.
The datavessel mobile app is live on Android today, so you can approve agent runs, ask questions about your numbers, and keep the loop moving from anywhere. The iOS app is coming soon — we’re in the final stretch of App Store review.
One step closer to the vision
The vision was never “another dashboard.” It was agents that do real work across your real stack, with you in control of what ships — reachable from wherever you happen to be working. Chat client, terminal, phone, IDE.
Every new surface is one step closer to that: the same agents, the same approvals, the same data — now in more of the places the work actually happens. The CLI is the latest step. Mobile on iOS is next.
You pilot. Agents do the work. And now they’re a npm i -g datavessel-cli away.
Explore the CLI at datavessel.io/cli, the MCP server at datavessel.io/mcp, or get started at app.datavessel.io.
Leave a Reply