Open standard · Model Context Protocol

Your data.
Wherever your AI works.

DashboardHC opens your operations to whichever AI your team already uses: Claude, ChatGPT, Copilot, and the next frontier models. Same OAuth, same scoping, no vendor chatbot to adopt.

See the setup Explore live

ChatGPT · DashboardHC

Where are my biggest areas to focus on with staffing?

get_data_model ask_data · labor ask_data · agency ask_data · ot get_dimension · building

Two patterns drive most of the labor variance.

Agency use at Willow Lane and Magnolia. Both running 11%+ agency hours, up from 4% last year. Most of the Mountain region's extra labor cost.
Weekend overtime at Sunset Ridge. Up 38% quarter over quarter, the biggest overtime driver in that building.

Willow Lane alone is about $48K of monthly variance. Want a memo for the regional?

Yes, draft the memo Show building detail

Why this matters

Your operational data should work everywhere.

MCP is just the open standard that makes it possible. The point is simpler: don't get trapped in someone else's AI silo. Your AI of choice can call real tools on your real data, on behalf of a real user, with real permissions.

PHI-safe by design

The MCP connector blocks direct patient identifiers at the query layer. Your AI never sees PHI, by design. The same permission model your team signs in with carries straight through to every AI call.

Role-based access

Your AI inherits the exact same permissions as the signed-in user: entity, region, measure, and time scoping. OAuth for people. Named, scoped API tokens for autonomous agents. Tokens revocable from the Integrations page.

Designed for multi-step analysis

Tools are atomic on purpose. The AI can plan, run multiple ask_data calls, and combine the results into one rich answer, whether driven by a person or a scheduled agent.

Available tools

A small surface area, on purpose.

You ask in plain English. The model picks the right tool and stitches the answer together.

get_data_model

Always called first. Returns measures, dimensions, buildings, and the P&L structure for this account, plus a routing guide so the AI picks the right tool for the question type.

get_dimension

Looks up exact member names within a dimension before they're used in a query. Drill-down and search modes for big dimensions like buildings, providers, or payers.

ask_data

Operational questions in plain English: occupancy, census, staffing, revenue trends. The AI generates and runs the underlying query, returns data plus an AI summary.

financial_review

Built for P&L, expenses, budget vs. actual, variance, and named GL line items. Use this when a question is fundamentally financial.

list_dashboards

Returns the dashboards available to a user, so the AI knows what visuals can be pulled before running one.

run_dashboard

Runs a specific DashboardHC dashboard for a given date and returns its chart data, ready for the AI to summarize or drop into a deck.

data_review

Runs all saved dashboard queries across multiple time periods (current month, prior, prior quarter, prior year) and returns a consolidated executive snapshot.

How multi-step analysis works. Ask "How does this month's census compare to last month by building?" and the AI calls ask_data for this month, then again for last month, then composes the comparison itself. For broad snapshots, data_review does the heavy lifting in a single call. For P&L-shaped questions the AI routes straight to financial_review.

Claude Desktop · Recommended

Connect Claude. Four steps.

Open the config file

In Claude Desktop, go to Settings → Developer → Edit Config.

Paste your DashboardHC server config

Replace the placeholders with the OAuth credentials your DashboardHC admin will issue you. (No admin yet? Email help@dashboardhc.com.)

json

{
  "mcpServers": {
    "dashboardhc": {
      "command": "npx",
      "args": ["mcp-remote", "https://apis.dashboardhc.com/v2/mcp"],
      "env": {
        "OAUTH_CLIENT_ID": "<your-client-id>",
        "OAUTH_CLIENT_SECRET": "<your-client-secret>"
      }
    }
  }
}

Restart & sign in

Restart Claude Desktop. The first time you call a DashboardHC tool, Claude opens a browser window for you to sign in with your DashboardHC credentials.

Start asking questions

Try: "What's occupancy for yesterday?" or "Build me a one-page operations review for last week."

ChatGPT · Custom GPT

Or wire up a Custom GPT.

ChatGPT supports DashboardHC as an Action on Custom GPTs, with the same OAuth model. Once configured, your team can share the GPT inside your workspace.

Create a Custom GPT

chatgpt.com → Explore GPTs → Create. Paste a starter prompt that points at the DashboardHC tools.

Add the action

Configure → Actions → Create new action. Import the OpenAPI schema URL we provide on signup.

Set up OAuth

Authentication → OAuth. Plug in your client ID, secret, and the DashboardHC authorization & token URLs from your Integrations page.

Save & chat

The first call prompts each user to sign in via OAuth. From then on, the GPT calls DashboardHC on their behalf.

Your DashboardHC admin can see every active OAuth connection in Integrations. Tokens can be named, audited, and revoked at any time.