MCP prompts and configuration resources
Read the project overview resource and request five built-in MCP prompts for setup, personalization, surveys, CTAs, and results pages.
Use MCP prompts to request setup, survey, and personalization instructions for your AI assistant. Read rightmessage://config/overview to discover your projects. The server provides five prompts and one resource. These requests return text, not tool calls that create, connect, or publish anything.
Examples below are illustrative and were not executed. They show MCP method names and parameters. Text responses are excerpts of returned content, not complete MCP transport envelopes.
Resource: rightmessage://config/overview
Discovers projects available to the authenticated user. Its registered name is config-overview, its title is Projects Overview, and its MIME type is text/plain. Use the URI with resources/read, not a tool named list-projects.
The required uri string is exactly rightmessage://config/overview. There are no project-selection, pagination, or filter arguments.
Illustrative text excerpt:
The resource lists projects belonging to the user that are eligible for MCP access. Expired fixed-end trials are excluded. It includes the authenticated user's name and email, public project IDs, configured domains, configuration counts, last publication date, and whether a draft exists. It uses draft configuration first, then published configuration. The offers count comes from the shared offers collection, which can include forms.
The segment groups count excludes dimensions linked to a library question; questions counts every library question, including inputs without a dimension. Those two counts are not a direct inventory of all dimensions.
Without authentication it returns No authenticated user. Please authenticate to view projects. With no eligible projects it returns No MCP-enabled projects found for the authenticated user. A project without configuration is labeled status: no configuration yet.
Reading the resource does not select an active project or alter configuration. Pass its ID explicitly to tools or use set-active-project. Treat the identity and project information as account data.
Prompt: personalization-assistant
Returns general instructions for an assistant planning and implementing RightMessage personalization. It covers project discovery, resource dependencies, and saving edits as drafts until publication is approved.
The goal string is required. This prompt has no project_id argument.
Illustrative response excerpt:
The returned guide tells the assistant how to discover project configuration and use related tools. The supplied goal is inserted as text. This request does not inspect a website or execute the plan. Always provide the required goal argument.
The prompt's statement that every question requires a segment group is outdated. That applies to choice questions, not input questions. Input questions use display_type: "input", an input_type, no dimension, and empty options. Follow the question tool contract rather than creating a group solely to satisfy that sentence.
Prompt: setup-wizard
Returns a setup workflow for a website, optionally tailored to an email platform.
| Argument | Type | Required | Meaning |
|---|---|---|---|
website_url | string | Yes | Website to discuss. |
platform | string | No | Email platform name for the setup guidance. |
Illustrative response excerpt:
The platform adds context to the returned instructions. The prompt does not validate credentials, connect an integration, create a project, crawl the URL, or install a snippet. Those are separate operations with their own permissions and effects. OAuth connections still require the dashboard.
Prompt: survey-builder
Returns instructions for planning and building a survey. The five phases gather context, interview you, present a strategy, build after approval, and hand off the draft.
The goal string is required. This prompt has no project ID or nested configuration arguments.
Illustrative response excerpt:
The response includes guidance on conversational questions, branching, scoring, outcome offers, and resource creation order. It asks for strategy approval before building and says to keep created content in draft. Requesting the prompt itself creates none of these resources.
Clients without prompt support can request the same guidance through get-survey-builder-guide in the setup tools reference. That tool makes its goal optional; the prompt declares it required.
Prompt: smart-cta-builder
Returns a strategy workflow for choosing targeted CTAs from existing visitor data, rather than adding questions by default.
The goal string is required. This prompt has no project_id argument.
Illustrative response excerpt:
The guide covers audience routing, offers, widget placement, strategy approval, and a draft handoff. Requesting it does not connect to your email platform, read visitor profiles, or build a flow. The corresponding tool fallback is get-smart-cta-builder-guide.
Prompt: quiz-results-page
Returns instructions for designing a results page around a quiz outcome dimension.
| Argument | Type | Required | Default and meaning |
|---|---|---|---|
quiz_segment_group_id | string | No | Empty by default; the assistant is told to discover the dimension. |
page_url | string | No | Empty by default; the assistant is told to ask or propose a URL. |
cta_type | string | No | Empty by default; examples include book_call, free_trial, download, and custom. |
page_builder | string | No | html; suggested values also include wordpress, webflow, and other. |
The suggested values are guidance, not enforced enums. Prompt arguments are strings, not flow nodes or executable code.
Illustrative excerpt showing one of the guide's substitution values:
The returned text substitutes the provided dimension and page details into a page-building guide. It does not create a page, change your CMS, visit the supplied URL, or publish RightMessage configuration. The corresponding fallback tool is get-quiz-results-page-guide.
Permissions, failures, and executable schemas
Prompts return instructions with your arguments inserted. They do not select a project or check a particular resource's ownership. Access to the MCP connection still requires authentication. Tools the assistant later calls perform their own project and permission checks. Unknown prompt names or malformed MCP requests are protocol errors, not empty project results.
The guides contain illustrative tool and condition examples, including some naming conventions used by the in-app assistant. Treat those as workflow guidance, not a guaranteed executable schema for the hosted MCP server. Use the current advertised tools/list schemas and the relevant tool reference before making a call. A guide's recommendation to create, connect, or publish something is not user authorization for that action.
The quiz-results guide also says phrases cannot render in campaign content changes. The current runtime does support phrases in RightMessage campaign content replacements; it does not scan arbitrary page HTML for phrase tokens. See phrase tools for that distinction.
All five prompt requests are text retrievals. They do not modify draft or live configuration. Later tool calls can have immediate account or provider effects even when website configuration remains unpublished.
For connection requirements and project access, see MCP authorization, project selection, and discovery.