How to Use the MCP Server Inside Ask AI
Ask AI now runs actions through the Sparkforge MCP (Master Control Panel) server. Type natural-language commands to create contacts, update opportunities, or bulk-tag — all permission-checked and logged.
Ask AI used to answer questions. With MCP, it now runs actions. Tell the chat to create a contact, reschedule an appointment, or bulk-update opportunities, and the Master Control Panel does the work — permission-checked, logged, and reversible where possible.
Why this matters
- Zero-click workflows: "Add Jane Smith to the NY Leads pipeline" — done.
- Context-to-action loop: ask for data, then act on it in the same thread ("Tag them as Hot").
- Universal coverage: anything the MCP API supports — contacts, payments, automations — runs in chat.
- Role-aware safety: permissions match your user role; no one exceeds their access.
- Audit and undo: every action logs to Activity Feed; destructive commands ask for confirmation.
How it works
- Parse intent. The AI identifies the verb (create, update, delete) and the object (contact, invoice).
- Generate payload. It maps your words to MCP fields like firstName or pipelineId.
- Permission check. The system verifies your role allows that action.
- Confirmation (destructive only). For delete or bulk edits, Ask AI shows a summary and asks "Proceed?"
- Execute and respond. MCP runs the call; success or error details return inline and log to the Audit Feed.
Command examples
| Action | Natural-language example | Behind the scenes |
|---|---|---|
| Create | "Create a contact named John Doe with phone 555-1234." | createContact() |
| Update | "Change pipeline stage of Opportunity #456 to Won." | updateOpportunity() |
| Query | "Show me tomorrow's booked appointments." | listAppointments() |
| Delete | "Delete coupon code SUMMER25." | deleteCoupon() — confirmation required |
| Bulk | "Tag all contacts from last week's webinar as Webinar-2025." | iterates updateContact() |
Tip: end any command with "and show me the result" to receive a JSON snippet for quick validation.
Safety and permissions
- Role enforcement: chat privileges mirror UI permissions — no back doors.
- Confirmation prompts: any irreversible change triggers a Yes/No step.
- Audit logs: find every AI-initiated action under Settings > Audit Logs > AI Chat.
- Credit usage: each successful MCP command costs 1 AI credit; previews and denied commands are free.
Prompt recipes
- Single-action: "Create invoice for $299 for contact john@demo.com due in 30 days."
- Two-step data + action: "List opportunities lost this week, then move them to Re-Engage stage."
- Conditional logic: "For contacts tagged Trial-Ending, if last login > 14 days ago, move to Churn Risk pipeline."
Frequently Asked Questions
How do I see the exact payload Ask AI sent to the server?
Click the timestamp on the chat reply to open the raw JSON payload in the Audit Log sidebar.
Can I undo a mistaken action?
For most actions, yes. Type "Undo last action" in the same chat, or use the Audit Log entry to reverse the change. Some destructive operations may not be reversible.
Does Ask AI respect my role permissions?
Yes. The MCP layer enforces the same permissions as the UI. If you cannot do an action manually, you cannot do it through Ask AI either.
How much does each command cost?
1 AI credit per successful MCP command. Previews and commands denied by permission checks are free.
Are bulk operations safe?
Bulk and delete operations always trigger a confirmation prompt with a summary of affected records. Nothing runs until you confirm.
Was this article helpful?
Thanks for the feedback.