VICIPanel's MCP server exposes 11 promoted tools (shown in AI client tool menus by default) plus 34 catalog actions (discovered on demand). Every tool respects the calling user's permission level, tools that require Manager access return a permission error for Agent-level tokens.
11 promoted tools
- whoami
- Return current auth context, who the AI is acting as and what scope the token has.
- health_check
- Ping the VICIPanel tenant and surface the status of key subsystems (dialer, database, carriers).
- list_campaigns
- List active campaigns with live KPIs, contact rate, drop rate, agent count.
- search_actions
- Fuzzy-search the catalog of 34 extended actions for one that matches your natural-language intent.
- execute_action
- Run a catalog action by ID with structured arguments. The general-purpose 'do a thing' tool.
- show_campaign_dashboard
- Render the live campaign dashboard widget inline in the AI client.
- show_agent_wallboard
- Render the agent status grid widget.
- show_cid_trace
- Render the CID decision cascade for a specific call ID, debug tool for compliance teams.
- show_did_inventory
- Render the DID inventory filtered by country, state, or campaign assignment.
- show_recording_player
- Render an inline audio player with transcript for a specific call recording.
- show_lead_detail
- Render a contact card for a specific lead with call history and dispositions.
34 catalog actions
Catalog actions cover every major operation: create/pause/resume campaigns, assign DIDs to CID groups, export CDRs, trigger CRM syncs, run reports, update lead dispositions, and more. The full list is discoverable at runtime, ask the AI 'what actions can you run on my VICIPanel?' and it will call search_actions for you.
Destructive actions require confirmation
Any action that modifies state (pausing campaigns, deleting leads, changing billing) requires the AI client to pass a confirm=true argument. This is enforced server-side, an AI client that forgets the confirmation gets an error, not an accident.