You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Payload CMS just shipped plugin-mcp (Model Context Protocol server), and they're actively bumping the SDK (v1.25.2 → v1.27.1 as of March 16, 2026). MCP is quickly becoming the standard protocol for AI tools (Claude Desktop, ChatGPT Plugins, Cursor, Windsurf, Zed) to interact with external services — including CMSes.
This is the next frontier of AI-first CMS: not just AI generating content inside the CMS, but AI tools being able to query, create, update, and publish content from outside the CMS, natively, through the agent's tool interface.
Numen has no MCP support today. Payload is already shipping it. We need to move fast.
What This Looks Like
An MCP server for Numen would expose the following tools to any MCP-compatible AI client:
Content Tools
numen_list_content — List content items with filters, pagination
numen_get_content — Fetch a content item by ID/slug
numen_create_content — Create a new content item (triggers AI pipeline optionally)
numen_update_content — Update existing content
numen_publish_content — Publish/unpublish a content item
numen_search_content — Semantic + keyword search across all content
Pipeline Tools
numen_run_pipeline — Trigger the AI content pipeline from a brief
numen_get_pipeline_status — Check pipeline run status
numen_get_pipeline_result — Fetch generated content from a pipeline run
Media Tools
numen_list_media — List media library items
numen_get_media — Fetch media metadata and URL
numen_upload_media — Upload media from URL or base64
Knowledge/Schema Tools
numen_list_content_types — List available content types
numen_get_content_type — Get schema for a content type
numen_query_knowledge_graph — Query the knowledge graph (when available)
Why This Is a Competitive Moat
Numen's core differentiator is AI-first architecture. MCP support is the natural extension of this:
Scenario
Today
With MCP
Claude writes a blog post
User copies text into Numen
Claude directly creates + runs pipeline
Cursor refactors docs
Manual copy-paste
Cursor queries Numen, updates, publishes
Windsurf agent updates copy
Zero integration
Native: read, edit, version, publish
CI/CD pipeline creates release notes
REST API integration
MCP tool call, no custom code
The key insight: Payload's MCP server means Payload users can say "Claude, update the homepage copy" and it just works. Numen should be able to say the same — but better, because our pipeline means Claude can say "Numen, write a blog post about X" and the entire pipeline (generate, illustrate, SEO, review, publish) executes autonomously.
reacted with thumbs up emoji reacted with thumbs down emoji reacted with laugh emoji reacted with hooray emoji reacted with confused emoji reacted with heart emoji reacted with rocket emoji reacted with eyes emoji
Uh oh!
There was an error while loading. Please reload this page.
-
The Opportunity
Payload CMS just shipped
plugin-mcp(Model Context Protocol server), and they're actively bumping the SDK (v1.25.2 → v1.27.1as of March 16, 2026). MCP is quickly becoming the standard protocol for AI tools (Claude Desktop, ChatGPT Plugins, Cursor, Windsurf, Zed) to interact with external services — including CMSes.This is the next frontier of AI-first CMS: not just AI generating content inside the CMS, but AI tools being able to query, create, update, and publish content from outside the CMS, natively, through the agent's tool interface.
Numen has no MCP support today. Payload is already shipping it. We need to move fast.
What This Looks Like
An MCP server for Numen would expose the following tools to any MCP-compatible AI client:
Content Tools
numen_list_content— List content items with filters, paginationnumen_get_content— Fetch a content item by ID/slugnumen_create_content— Create a new content item (triggers AI pipeline optionally)numen_update_content— Update existing contentnumen_publish_content— Publish/unpublish a content itemnumen_search_content— Semantic + keyword search across all contentPipeline Tools
numen_run_pipeline— Trigger the AI content pipeline from a briefnumen_get_pipeline_status— Check pipeline run statusnumen_get_pipeline_result— Fetch generated content from a pipeline runMedia Tools
numen_list_media— List media library itemsnumen_get_media— Fetch media metadata and URLnumen_upload_media— Upload media from URL or base64Knowledge/Schema Tools
numen_list_content_types— List available content typesnumen_get_content_type— Get schema for a content typenumen_query_knowledge_graph— Query the knowledge graph (when available)Why This Is a Competitive Moat
Numen's core differentiator is AI-first architecture. MCP support is the natural extension of this:
The key insight: Payload's MCP server means Payload users can say "Claude, update the homepage copy" and it just works. Numen should be able to say the same — but better, because our pipeline means Claude can say "Numen, write a blog post about X" and the entire pipeline (generate, illustrate, SEO, review, publish) executes autonomously.
Competitor Context
plugin-mcpshipped, actively maintained (SDK bumped March 16, 2026)Numen can be the first AI-native CMS with a full pipeline-aware MCP server. Not just CRUD tools, but tools that trigger multi-agent workflows.
Implementation Scope
Phase 1 (M): MCP server exposing content CRUD + search tools via
@modelcontextprotocol/sdknumen mcp:serve(stdio transport for Claude Desktop)GET /mcpwith SSE for remote clientsPhase 2 (S): Pipeline tools —
numen_run_pipeline,numen_get_pipeline_statusPhase 3 (S): Knowledge graph + semantic search tools
Proposed Acceptance Criteria
numen mcp:servestarts an MCP stdio server/api/mcpwith SSEBeta Was this translation helpful? Give feedback.
All reactions