Let an AI agent build your KPI dashboard

Setting up a metrics wall used to mean clicking through a dashboard yourself: create a board, add tiles, wire up webhooks, push a first value to see it work. We just built a shortcut, and it ships with Dashwall from day one: tell an AI agent what you want to track, and it builds the wall for you. Dashwall comes with an MCP server, so agents like Claude can create boards, add tiles, push live values and hand you the display link, all from a single prompt.

What's behind it: MCP

MCP (Model Context Protocol) is the open standard that lets AI assistants use external tools, the same mechanism Claude uses to read files or query databases. The Dashwall MCP server exposes your boards as a small set of tools an agent can call: create a board, list boards, add or edit tiles, push a metric value.

There are two ways to let an agent in. The hosted server speaks OAuth: the client registers itself, sends you to a Dashwall screen naming the application and what it is asking for, and gets its own token once you approve, with nothing to copy. Or you create an API key on the Agents page, which can be read-only and can be given an expiry. Either way your password stays yours, and access can be withdrawn at any time.

From prompt to wall in three steps

1. Connect the MCP server. For Claude Code it's one command: claude mcp add --transport http dashwall https://mcp.dashwall.io/mcp, then approve it in the browser. Claude Desktop and any other MCP client work too.
2. Or run it locally with a key you hold yourself: claude mcp add dashwall -e DASHWALL_API_KEY=dw_... -- npx -y dashwall-mcp.
3. Ask for a dashboard. "Set up a board for our support team: open tickets, first-response time, CSAT", and the agent creates the board and the tiles, pushes a first value so you can see it move, and replies with the display link. Open that link on the office TV and you're done.

The display link the agent hands back is the same public board URL described on the how it works page. The agent watches the result of its own work the same way you do.

Agents don't just build walls, they feed them

The setup story is nice, but the more interesting part is what happens after. Every tile still has its plain webhook URL, and the MCP server can push values too. That opens up patterns that are awkward with hand-built dashboards:

• An agent that reports its own progress. A long-running agent such as a nightly data job, a migration or a scraper can push "records processed" to a tile while it works, and the whole team watches it climb on the wall.
• Metrics without an integration. Ask an agent to look something up every morning and post it: yesterday's signups, open PRs, app-store rating. No Zapier scenario, no cron job to maintain.
• Walls that evolve. "Add a tile for the new product line and set a milestone at 100 sales" is a sentence, not a settings session. When the milestone hits, the wall celebrates with confetti. That part we kept.

If you're still deciding what belongs on your wall in the first place, the use cases page has per-team examples, and how to show live metrics on an office TV covers the screen-side setup.

Try it

Dashwall is live, and agents have been part of it from day one: create a free account and the first thing you can do with it is hand a key to Claude and say "build me a dashboard".