Open Elements Content MCP

An MCP server that crawls, indexes, and makes searchable/retrievable the content (primarily blog posts) of several Open-Elements-adjacent websites — exposed to AI agents over the standard /mcp streamable-HTTP endpoint.

The application is standalone and built on spring-services as a library: it reuses that library's MCP server and Meilisearch building blocks and adds the content crawling/indexing/search pipeline in the com.openelements.content package.

Status: project skeleton (spec 001-project-skeleton). The app boots and exposes the /mcp endpoint with no content-specific tools yet. The crawler, indexer, search service, and MCP tools arrive in later specs — see docs/specs/INDEX.md and docs/roadmap.md.

Prerequisites

• Java 21 (enforced by the Maven parent)
• Maven 3.9.11+
• Meilisearch (optional for boot) — needed for actual indexing/search. The app tolerates an unreachable Meilisearch at startup and comes up anyway.

Build

mvn clean package

This runs the tests and produces an executable jar in target/.

The build depends on spring-services:1.3.0-SNAPSHOT, resolved from the Central Portal snapshot repository declared in pom.xml.

Run

java -jar target/content-mcp-0.1.0-SNAPSHOT.jar

The app starts on port 8080 and exposes /mcp. It connects to Meilisearch at http://localhost:7700 by default (override via environment variables below).

Test

mvn test

Configuration

Configuration lives in src/main/resources/application.yaml. Common overrides (all via environment variables):

Variable	Default	Purpose
MEILISEARCH_HOST	http://localhost:7700	Meilisearch base URL
MEILISEARCH_MASTER_KEY	(empty)	Meilisearch master key
MEILISEARCH_INDEX_PREFIX	content_	Index name prefix
SPRING_DATASOURCE_URL	in-memory H2	Datasource for the library's api-key/user tables
SPRING_DATASOURCE_USERNAME	sa	Datasource user
SPRING_DATASOURCE_PASSWORD	(empty)	Datasource password
OAUTH2_JWK_SET_URI	placeholder	JWK set URI for the library's OAuth2 resource server

Notes on the wiring

spring-services couples its MCP server to a JPA-backed api-key/user stack and depends on spring-boot-starter-data-jpa non-optionally, so a datasource is mandatory. This app supplies an embedded H2 database that backs only those library tables — the content pipeline itself keeps no relational state (its store is Meilisearch). The /mcp endpoint is secured by the library defaults; real scoped-key authentication is added in a later spec (013). See docs/specs/001-project-skeleton/design.md for the full rationale.

Documentation

• docs/content-mcp-technical-design.md — overall technical design
• docs/roadmap.md — implementation roadmap
• docs/specs/ — per-step specifications