Claude Code Integration

d1-eloquent ships with first-class Claude Code support through two complementary tools: a skill that teaches Claude how to write correct d1-eloquent code, and an MCP server that gives Claude live access to your D1 database.

Skill

The d1-eloquent skill is a set of reference documents that Claude Code loads automatically when it detects you're working with d1-eloquent. It covers model definitions, migrations, query building, relationships, hooks, casting, seeders, factories, and every CLI command.

When the skill is active, Claude will:

• Generate models with correct BaseModel<TAttrs> patterns, TEXT primary keys, and proper casts
• Write migrations using the schema builder (not raw SQL)
• Use declarative static relations instead of legacy helpers
• Apply configure(env) auto-wiring or explicit db arguments consistently
• Follow all d1-eloquent conventions (soft deletes, revision tracking, mass assignment, etc.)

Install the Skill

# Clone into your Claude Code skills directory
git clone https://github.com/Orphnet/d1-eloquent-skill.git ~/.claude/skills/d1-eloquent-skill

That's it. Claude Code discovers skills from ~/.claude/skills/ automatically. The skill activates whenever you mention d1-eloquent, Cloudflare D1, or ask Claude to scaffold a Workers + D1 project.

No configuration needed

The skill has no dependencies or build step. It's a set of markdown reference files that Claude reads on demand.

MCP Server

The MCP server gives Claude (and any MCP-compatible client) live read/write access to your D1 database. Claude can query records, inspect schema, create and update data, browse revision history, and time-travel — all through natural language.

The server understands your d1-eloquent models: it respects soft deletes, writes audit revisions, applies the correct primary keys, and auto-generates UUIDs.

Install in Claude Code

claude mcp add d1-eloquent -- npx @orphnet/d1-eloquent-mcp

Install in Claude Desktop

Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "d1-eloquent": {
      "command": "npx",
      "args": ["@orphnet/d1-eloquent-mcp"],
      "env": {
        "D1_BINDING": "DB",
        "MODELS_PATH": "./src/models"
      }
    }
  }
}

Configuration

Variable	Default	Description
WRANGLER_CONFIG	./wrangler.jsonc	Path to wrangler config
D1_BINDING	DB	D1 binding name
D1_LOCAL	true	Use local D1 (false for remote)
MODELS_PATH	./src/models	Where model classes live
D1_REMOTE_ACCOUNT_ID	--	Cloudflare account ID (remote mode)
D1_REMOTE_DATABASE_ID	--	D1 database ID (remote mode)
CLOUDFLARE_API_TOKEN	--	API token (remote mode)
MAX_ROW_LIMIT	1000	Maximum rows returned per query

Available Tools

Schema & Inspection

• d1_list_tables -- list all tables with column info
• d1_describe_table -- full schema, indexes, foreign keys, row count
• d1_migration_status -- applied vs pending migrations

Querying

• d1_query -- read-only SQL with parameter binding
• d1_find -- ORM-style filters (no SQL needed)
• d1_find_by_id -- find by primary key
• d1_count -- count with optional filters

Mutations

• d1_create -- insert with auto-UUID and timestamp generation
• d1_update -- update by primary key (dirty tracking)
• d1_delete -- delete or soft-delete
• d1_restore -- restore soft-deleted records

Revisions & Time-Travel

• d1_revisions -- revision history for a record
• d1_as_of -- reconstruct state at a past timestamp
• d1_revert -- roll back to a specific revision

Utilities

• d1_bulk_create -- insert up to 100 records at once
• d1_raw_query -- arbitrary SQL (writes require confirmation)

Safety

• d1_query only allows SELECT / WITH / EXPLAIN / PRAGMA
• d1_raw_query requires explicit confirm_write: true for mutations
• All queries enforce a configurable row limit (default 1000)
• Table and column names are validated against injection patterns
• All queries use parameterized binding

Resources

• d1-eloquent-skill on GitHub -- skill source and reference docs
• Claude Code documentation -- skills, MCP servers, and configuration