MCP server: Measured in your own LLM

Measured ships a native Model Context Protocol server. Point Claude Desktop, Claude Code, Cursor, or any MCP-capable client at it and your LLM can operate Measured directly — list sources, profile assets, author rules, trigger validations, and triage failures — all in plain language, from the tool you already work in.

No plugin to install on our side. It's a hosted Streamable HTTP endpoint plus a personal API key.

What your LLM can do over MCP

  • See your estate — data sources, assets, schemas, profile stats, health.
  • Profile on demand — kick off a profile run and read back per-column statistics.
  • Author rules — draft and update rules against real profile data, browse the rule gallery, and use the server-side AI tools (mc_ai_suggest_rules, mc_ai_profile_summary) to propose validated rule sets you approve.
  • Run validations — trigger a validation, poll the run, and summarise pass/fail with per-rule detail.
  • Triage — list failures and anomalies, acknowledge anomalies, create alert rules and schedules.
  • Answer questions — mc_search_docs retrieves this documentation, so "how do I connect Snowflake?" gets a sourced answer inside your editor.

37 tools, every name prefixed mc_ — but you never call them by name. You ask; the model picks the tools.

Connect in two minutes

  1. Mint a personal API key in the app: Settings → API Keys → New API key. Copy the dqk_… value (shown once). The key acts as you — it carries your role and optional workspace scope, and the API enforces the same org/workspace/role isolation as always.

  2. Add the server to your client. With Claude Code:

claude mcp add --transport http measured https://mcp.measured.cloud/mcp \
  --header "Authorization: Bearer dqk_..."

Or in any client's MCP config JSON (Claude Desktop, Cursor, and friends):

{
  "mcpServers": {
    "measured": {
      "type": "http",
      "url": "https://mcp.measured.cloud/mcp",
      "headers": { "Authorization": "Bearer dqk_..." }
    }
  }
}

Self-hosting? The endpoint is https://<your-mcp-host>/mcp — the exact URL (with copy-paste config) is shown in Settings → AI in your deployment.

  1. Say hello. Ask "Who am I connected as?" — the model calls mc_whoami and reports your user, org, workspace scope, and role. That confirms auth end to end.

Things to try

"List my data sources and tell me which assets haven't been profiled this week."

"Profile ASSET_MGMT.PORTFOLIO_POSITIONS and suggest rules from the results. Show me before you save anything."

"Run the nightly validation and explain any failures with the offending columns."

"Any unacknowledged anomalies? Acknowledge the low-severity ones and summarise the rest."

Security notes

  • Scope the key. Keys can be limited to one workspace; a scoped key can't see or touch anything outside it.
  • Store it like a password — in your client's secret store, never in a checked-in config file.
  • Revoke instantly. Delete the key in Settings → API Keys and the next tool call fails with an auth error.
  • Keep everything on your network. On Enterprise, the MCP server and all AI features run against an on-prem or self-hosted LLM via the Privacy Agent — prompts, credentials, and data never leave your environment.

Tool catalogue (grouped)

  • Identity — mc_whoami
  • Data sources — mc_datasource_list · mc_datasource_get · mc_datasource_create · mc_datasource_test · mc_datasource_discover_assets
  • Data assets — mc_asset_list · mc_asset_get · mc_asset_create
  • Profiling — mc_profile_run · mc_profile_get
  • Rules — mc_rule_suite_list · mc_rule_suite_get · mc_rule_suite_create · mc_rule_suite_update · mc_rule_type_gallery
  • Validations & runs — mc_checkpoint_list · mc_checkpoint_get · mc_checkpoint_create · mc_checkpoint_run · mc_run_get
  • Results — mc_result_list · mc_result_get
  • Anomalies — mc_anomaly_list · mc_anomaly_acknowledge
  • AI assist — mc_ai_suggest_rules · mc_ai_profile_summary · mc_rule_suggestion_context · mc_suggestion_list · mc_suggestion_get · mc_suggestion_review · mc_suggestion_accept
  • Ops — mc_schedule_list · mc_schedule_create · mc_alert_rule_list · mc_alert_rule_create
  • Docs — mc_search_docs