Tools, resources and prompts

tools/list, resources/list and prompts/list return only what the credential can use right now. When a grant changes, the next list reflects it.

Tools

Four discovery tools are always available:

Tool	Returns
`gateway_context`	The installation, its scopes, data access, restrictions and expiry, and the available guides
`guide_read`	One guide written for agents
`websites_list`	The granted websites. Needs `website_read`.
`capabilities_list`	The operations available on one website

Every other tool has the same name, inputs and permissions as a REST operation. The all operations page shows the MCP tool for each one.

Successful results have structuredContent: {"data": <REST response body>} and the same JSON as text.
Failures have isError: true and structuredContent: {"error": {"code", "status", "request_id", "retry_after"}}.
Clients that negotiate 2025-06-18 or later receive output schemas for the main read tools.

guide_read and the matching inside://gateway/{guide} resources return Markdown written for agents. Only guides the credential can use are listed.

Guide	Covers
`overview`	Current tools, data boundaries and safe retries
`investigation`	Reading conversations and transcripts
`contacts`	Contact lookups, linked history and notes
`reporting`	The AI conversation summary report
`configuration`	AI feature metadata and previews
`agent_tasks`	Preparing A2A tasks
`event_recovery`	Following and recovering event subscriptions

Each template is listed only when its permissions are granted.

inside://gateway/websites/{website_id}/capabilities
inside://gateway/conversations/{conversation_id} and .../contacts
inside://gateway/contacts/{contact_id}, .../visitors and .../notes
inside://gateway/websites/{website_id}/ai-conversation-summary/{range}, where range is today, last-7-days or last-month
inside://gateway/websites/{website_id}/ai-conversation-summary/date/{start}/{end}
inside://gateway/websites/{website_id}/settings/schema/ai/features, .../settings/ai/features and .../settings/ai/features/{feature_id}

Prompts prepare structured requests. Getting a prompt doesn’t read customer data or change anything.

Prompt	Prepares
`customer_themes`, `product_questions`	A review of 1 to 20 explicit conversations
`performance_summary`	An AI conversation summary for one website and period
`contact_investigation`	Reads for one contact, optionally with history and notes
`configuration_preview`	An AI feature name or enabled-state preview
`support_insights_task`, `conversation_evidence_task`, `configuration_inspection_task`, `contact_inspection_task`, `offline_language_detection_task`	A ready-to-send A2A request