Persona: Consultant (Onboard, Govern, Isolate)

This page is a practical workflow for consultants who use MCP Engine across multiple customers/models and need repeatable onboarding, safe operating patterns, and governance awareness.

What you typically want

  • Onboard fast: model tour + “what matters” inventory + known risks.
  • Establish safe prompting conventions with the customer team.
  • Ensure governance expectations (mode/policy/license) are understood.
  • Avoid “cross-engagement bleed” (preferences/policy/tests/history shared on the server).

Consultant onboarding workflow (first session)

  1. Confirm environment constraints

    “What mode is this server in (full/read-only/browse-only) and what does that block?”“What license tier is active, and which Pro/Enterprise tools are enabled here?”“Is Service (XMLA) enabled in this deployment?”
  2. Connect intentionally and confirm context

    "List available models/datasets and ask me which one to connect to (don't auto-select). Confirm Desktop vs Service." For the full connection workflow, see Connecting and Context.

  3. Produce a “model tour” deliverable

    “Generate a 1–2 page model tour: purpose, main tables, key measures, relationships, calc groups, roles, and known risks.”
  4. Establish operating prompts (team baseline)

    Agree on:

    • plan-first + confirm-before-write,
    • small outputs (aggregates/topN),
    • masking when sharing outputs,
    • external formatter consent stance. Copy/paste: “Plan first. Don’t apply writes until I explicitly confirm with ‘yes’. Keep outputs small and masked.”

Isolation and shared-state gotchas (important)

Governance deliverables you can produce

  • Mode + tool availability summary for stakeholders.
  • A safe prompting “working agreement” for the team.
  • Impact analysis notes for planned refactors.
  • Validation query suite for go-live.

Risk & governance reminders

See also