@orphnet/d1-eloquent-mcp
@orphnet/d1-eloquent-mcp
Claude(및 Model Context Protocol 호환 클라이언트)에게 사용자의 d1-eloquent 프로젝트(모델, 마이그레이션, 스키마, 시더, 팩토리 및 CLI)에 대한 실시간 컨텍스트 인식 액세스를 제공하는 MCP 서버입니다.
bunx @orphnet/d1-eloquent-mcpClaude에게 제공되는 기능
도구 24개, 리소스 6개, 프롬프트 2개.
인트로스펙션(읽기 전용)
list_models·read_model·list_migrations·read_migration·migration_statusinspect_schema-sqlite_master를 쿼리하여 D1에 실제로 무엇이 있는지 확인합니다.query_d1- 로컬 D1에 대한 읽기 전용 SQL(SELECT/WITH/EXPLAIN/PRAGMA만 허용)list_seeders·list_factories·read_file·refresh_projectvalidate_model·validate_all- 선언된 모델 형태와 실제 스키마를 비교하여 드리프트(불일치)를 보고합니다.
생성 (d1-eloquent CLI에 위임)
make_model·make_migration·make_seeder·make_factorymake_resource- 모델 + 마이그레이션 + 시더 + 팩토리를 한 번에 생성make_pivot
실행 (d1-eloquent CLI에 위임, 기본적으로 로컬 D1 사용)
run_migrate·run_rollback·run_fresh·run_seedrun_generate- 스키마 차이(schema-diff): 모델과 실제 스키마 사이의 차이를 좁히는 마이그레이션을 작성합니다.
리소스 - Claude가 배경 컨텍스트로 가져올 수 있습니다.
d1-eloquent://project/summary- 프로젝트의 발견된 상태d1-eloquent://docs/api/base-modeld1-eloquent://docs/api/query-builderd1-eloquent://docs/api/relationshipsd1-eloquent://docs/api/transactionsd1-eloquent://docs/guide/quick-start
프롬프트 - 사전 구축된 다단계 워크플로
create-model- 한 번의 채팅에서 스캐폴드 + 마이그레이션 + 검증audit-schema- 모든 모델을 실제 D1 스키마와 비교하고 드리프트를 보고합니다.
Related MCP server: MCP Database Bridge
설치
Claude Code
claude mcp add d1-eloquent -- bunx @orphnet/d1-eloquent-mcpClaude Desktop
~/Library/Application Support/Claude/claude_desktop_config.json (macOS) 또는 %APPDATA%\Claude\claude_desktop_config.json (Windows):
{
"mcpServers": {
"d1-eloquent": {
"command": "bunx",
"args": ["@orphnet/d1-eloquent-mcp"]
}
}
}프로젝트별 (Claude Code)
.claude/mcp.json:
{
"mcpServers": {
"d1-eloquent": {
"command": "bunx",
"args": ["@orphnet/d1-eloquent-mcp"]
}
}
}현재 디렉터리가 아닌 프로젝트를 지정하려면:
{
"mcpServers": {
"d1-eloquent": {
"command": "bunx",
"args": ["@orphnet/d1-eloquent-mcp", "--project", "/path/to/your/d1-eloquent-project"]
}
}
}요구 사항
Node.js ≥ 22
wrangler가 설치되어 있고 D1 액세스를 위해 인증되어 있어야 합니다 (bun add -D wrangler)기존
@orphnet/d1-eloquent프로젝트 - 서버가 파일 트리를 스캔하여 모델/마이그레이션/시더를 발견합니다.
안전
query_d1은SELECT/WITH/EXPLAIN/PRAGMA외의 다른 쿼리를 거부하며, 다중 문장 페이로드도 거부합니다.read_file은..를 포함하거나/로 시작하는 경로를 거부합니다.생성 도구는 기존 d1-eloquent CLI를 호출합니다 - 자체 작성기를 포함하지 않습니다.
실행 도구(
run_migrate/run_rollback/run_fresh/run_seed)는 기본적으로 로컬 D1에서만 작동합니다. 원격의 경우 CLI를 통해 직접 실행하세요.
라이선스
MIT
Available Tools
24 toolsinspect_schemaA
Query the local D1 sqlite_master to list every table + the column definitions actually applied. Read-only.
Inputs: db, table.
list_factoriesA
List discovered factory files. Read-only.
Inputs: no named parameters.
list_migrationsB
List all discovered migration files with timestamps. Read-only.
Inputs: no named parameters.
list_modelsA
List all discovered BaseModel subclasses in the project, with table names and feature flags (soft deletes, revisions, relationships, casts). Read-only.
Inputs: no named parameters.
list_seedersA
List discovered seeder files. Read-only.
Inputs: no named parameters.
make_factoryB
Scaffold a new factory. Delegates to d1-eloquent make:factory.
Inputs: name (required).
make_migrationB
Scaffold a new migration. Delegates to d1-eloquent make:migration.
Inputs: name (required).
make_modelB
Scaffold a new model file. Delegates to d1-eloquent make:model.
Inputs: name (required), softDeletes.
make_pivotB
Scaffold a pivot table migration. Delegates to d1-eloquent make:pivot.
Inputs: table (required).
make_resourceB
Scaffold model + migration + seeder + factory in one go. Delegates to d1-eloquent make:resource.
Inputs: name (required), softDeletes.
make_seederC
Scaffold a new seeder. Delegates to d1-eloquent make:seeder.
Inputs: name (required).
migration_statusA
Run d1-eloquent status against the local D1 - shows which migrations have run vs pending. Read-only. The CLI resolves the D1 binding from wrangler config automatically.
Inputs: no named parameters.
query_d1A
Run a read-only SQL query (SELECT / WITH / EXPLAIN / PRAGMA only) against the local D1. Multi-statement payloads and writes are rejected.
Inputs: sql (required), db.
read_fileA
Read any project file by path (relative to project root). Read-only.
Inputs: path (required).
read_migrationB
Return the source of a migration file. Read-only.
Inputs: name (required).
read_modelA
Return the full source code of a model file plus parsed metadata (relations, casts, static config). Read-only.
Inputs: name (required).
refresh_projectB
Re-run project discovery, update cached state.
Inputs: no named parameters.
run_freshA
Drop all tables and re-run every migration. Delegates to d1-eloquent fresh --force.
Inputs: no named parameters.
run_generateA
Diff models against the project's migrations (schema diff) and report what a reconciling migration would change. Dry-run by default: prints per-model added/dropped/type-changed columns and writes NOTHING. Pass write=true to emit migration file(s) into the project's migrations directory (review them, then run_migrate). Delegates to d1-eloquent generate.
Inputs: name, model, write.
run_migrateB
Apply pending migrations. Delegates to d1-eloquent migrate against local D1.
Inputs: no named parameters.
run_rollbackA
Roll back the last migration batch. Delegates to d1-eloquent rollback.
Inputs: no named parameters.
run_seedC
Run seeders. Delegates to d1-eloquent seed.
Inputs: seeder.
validate_allA
Run validate_model across every discovered model. Returns per-model issues and a summary count of ok / warn / fail.
Inputs: no named parameters.
validate_modelA
Compare a single model's declared shape (table, casts, softDeletes, timestamps, revisions) against the actual local D1 schema and return any drift as a list of issues.
Inputs: name (required).
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 24 tools
Each tool targets a distinct resource and action (list vs read vs make vs run vs validate). Even closely related tools like validate_model and validate_all are clearly differentiated by scope (single vs all). There is no ambiguity in purpose.
The overwhelming majority use verb_noun snake_case (list_*, make_*, run_*, read_*). The only deviation is migration_status, which uses noun_noun instead of a verb prefix, but this is a minor inconsistency that does not hinder readability.
At 24 tools, the set falls in the 16-25 range that feels heavy. While each tool serves a distinct purpose in the D1/Eloquent workflow, the high number may be overwhelming and includes some conveniences (like make_resource) that could be assembled from others.
The tool set covers the full lifecycle: listing/reading artifacts, scaffolding (model, migration, seeder, factory, pivot), running migrations/seeds/rollbacks, schema inspection, validation, and even raw SQL queries. No significant gaps are apparent for the stated scope.
Maintenance
Related MCP Connectors
Hosted SQL Server MCP server for Claude, ChatGPT and Cursor, with each action switched per account.
Build, deploy and manage MaShop e-commerce projects from Claude, Cursor or any MCP client.
Hosted PostgreSQL MCP server for Claude, ChatGPT and Cursor, with each action switched per account.
Work on your own computers from Claude or ChatGPT: run commands, edit and search files.
Related MCP Servers
AlicenseAqualityCmaintenanceEnables AI-assisted analysis and optimization of Cloudflare D1 databases through comprehensive schema introspection, relationship mapping, validation, and optimization recommendations. Demonstrates semantic intent patterns with hexagonal architecture for maintainable database development workflows.43MIT- FlicenseNot gradedqualityCmaintenanceA secure, read-only MCP server that empowers Claude Desktop and AI agents to safely query and inspect local SQLite databases.-
- AlicenseNot gradedqualityBmaintenanceA Node.js MCP server that gives Claude Code direct access to file operations, database queries, PHP execution, SFTP deployment, and over 60 specialized tools for development and debugging.1MIT
- AlicenseAqualityCmaintenanceMCP server for Cloudflare D1 database introspection, providing schema analysis, relationship extraction, validation, and optimization suggestions via Semantic Intent patterns.52MIT