Skip to main content
Each prompt below is one paragraph you can paste into Claude Desktop, Cursor, or any MCP-enabled client. Replace demo-acme / alex-smith with your own org / rep ids.

Verify connectivity

Expected: echo_ping returns { pong: true, key_environment: "sandbox", ... } in under a second.

The hero workflow — analyze, classify, coach

What it exercises: parallel async tool calls, persona meta-analysis, Scout multi-turn chat. ~60 seconds end to end. This is the demo to show stakeholders.

Coach a single rep

What it exercises: generate_rep_synthesis with lookback_count=5 → returns synthesis with action plan citing specific calls.

Tag outcomes for win/loss analysis

What it exercises: tag_disposition, list_dispositions with date range. RevOps-flavored use case.

AI-generate a playbook from raw notes

What it exercises: upload_playbook_source (×2), generate_playbook_draft (async, ~20s), publish_playbook_draft. The wow moment.

Configure custom scoring criteria

What it exercises: create_custom_prompt (×3), list_custom_prompts. Sales-leader admin flow.

Reorder prompts after creation

What it exercises: list_custom_prompts then reorder_custom_prompts with the right ids array. Tests model’s ability to chain identifiers.

Quick rep snapshot

What it exercises: get_rep_stats with custom window + interpretation by the model.

Manual persona override

What it exercises: set_rep_persona_manually, get_rep_persona. Appends to history with source: "manual".

Scout deep-dive on a single call

What it exercises: get_scout_welcome_questions, ask_scout_about_call (multi-turn — Scout remembers turn 1 in turn 2).

Audit configuration

What it exercises: get_org, list_reps, list_playbooks active_only=true, list_custom_prompts, get_latest_org_insights (Phase F). Tests parallel-call orchestration.

Generate an org-level insights snapshot

What it exercises: generate_org_insights (async, ~30–60s, ~0.020.02–0.05), then the model parses the snapshot. Sales-leader monthly-review use case.

Leaderboard — pure SQL, no AI cost

What it exercises: get_leaderboard called three times with different sort_by values, then cross-referenced. No AI cost — pure SQL.

Set up + test a webhook

What it exercises: register_webhook (signing_secret returned ONCE), test_webhook. After this, rotate_webhook_secret is the only way to ever get a new secret.

Rotate a leaked webhook secret

What it exercises: rotate_webhook_secret. The model should also explain to update the verification code FIRST, then rotate, since the old secret stops working immediately.

Discoverability — “what can you do?”

The model will list all 60 tools grouped by domain, and (because every tool description has a cost callout) it’ll correctly tag the AI-cost ones.

Cleanup after demo

Soft-deletes the entire demo workspace. Reversible manually server-side.