Skip to main content
These playbooks are prompts you can give an MCP client (Claude, Cursor, etc.) connected to Oviond. Each lists the tools the agent will reach for and any safety note. Tool names and the full catalog are on Available tools.
Every call is automatically scoped to your authenticated account — you never pass an account ID, and the agent can only see your own clients and projects.

1. List and inspect clients

“List my Oviond clients grouped by folder, and tell me which ones have no active projects.”
Tools: clients_list, folders_list, projects_list, search What happens: the agent pages through your clients, groups them by folder, and cross-references each client’s projects. Read-only — nothing changes.

2. Create a client with a report (from a template)

“Create a client called Acme Demo with a Google Ads dashboard.”
Tools: templates_list (find the template) → clients_create_with_project What happens: the agent finds the matching template with templates_list (filter datasources=gadw, type=DASHBOARD → “Google Ads Dashboard”), then clients_create_with_project creates the client and clones that template into the first project — pages and widgets included. The create step is smart about the project:
  • “…with a Google dashboard” → the agent passes the matched template_id.
  • “…with a blank report” → pass blank: true for an empty project.
  • “…with a report” (unspecified) → a matching Starter template is cloned automatically.
To add a template-based report to an existing client, use projects_create_from_template.
This creates real records in your account. Confirm the client and report names before the agent runs it.

3. Duplicate a report for another client

“Duplicate the ‘Monthly Marketing Overview’ report for the client Beta Corp.”
Tools: search / projects_listprojects_duplicateprojects_update What happens: the agent finds the source report, duplicates it, and re-points the copy at the new client. Layout and widgets carry over.

4. Add a saved section to a report or dashboard

“Add my ‘Executive Summary’ section to the Acme dashboard.”
Tools: assets_list (find the section) → projects_add_section What happens: the agent finds the saved section with assets_list (search by name), then projects_add_section places its widgets on the project. A single-page project (typical dashboard) uses its one page automatically; a multi-page report returns its pages so the agent can pick the right page_id.

5. Apply branding

“Set Acme Demo’s brand color to #1E88E5 and generate a matching palette.”
Tools: branding_get, branding_update, company_generate_palette What happens: the agent reads current branding, generates a palette, and updates the logo/colors. Review the palette before applying.

6. Query live data

“For Acme, pull sessions and conversions from GA4 for the last 30 days, broken down by day.”
Tools: data_list_datasourcesdata_list_metrics / data_list_dimensions (discover ids), then data_query (or reports_get_metric_values for headline totals) What happens: the agent finds the datasource_id, discovers the valid metric/dimension ids, then queries live values — no guessing at any step. Read-only.
Metric and dimension identifiers are datasource-specific. If a query returns nothing, run data_test_connection — the client’s OAuth token may have expired and need reconnecting in the app.

7. Audit a report’s health

“Audit the ‘Monthly Marketing Overview’ report — what data sources are connected and are any automations paused?”
Tools: projects_audit_health (one call), or projects_get + datasources_list + automations_list What happens: the projects_audit_health workflow tool gathers the connected data sources and automation state in one step and flags likely issues (nothing connected, paused automations). Read-only.

8. Schedule a monthly report (with confirmation)

“Schedule the Acme monthly report to email on the 1st of each month to [email protected] — but show me the settings and ask before you activate it.”
Tools: automations_create, automations_test, automations_pause / automations_unpause, automations_history What happens: the agent drafts the automation (recipient, cadence, timezone), can send a test, and should leave it paused until you confirm activation.
This changes delivery state and sends real email. Ask the agent to create it paused and send a test to yourself first.

What MCP can’t do yet

  • Connect a data source. Connecting a source needs interactive OAuth, done once in the app. MCP can list, rename, and delete existing connections — and query the data from them — but it can’t create the connection itself.