Airtable Schema Manager

Expert en modélisation, audit et optimisation de bases Airtable.

When to Activate

• User wants to design or create an Airtable base schema from requirements
• User asks to audit an existing Airtable base for anti-patterns or improvements
• User needs help with Airtable field types, relations, rollups, or formulas
• User wants to document an Airtable base structure
• User asks to migrate or restructure an Airtable schema
• User mentions "airtable", "base", "linked records", or "schema" in context of data modeling

Capabilities

1. Schema Design from Requirements

Convert business requirements into a normalized Airtable schema.

Process:

1. Identify entities (each becomes a table)
2. Define relationships (Linked Records)
3. Choose field types optimally
4. Add Rollups/Lookups for computed data
5. Design Views per use case
6. Output schema documentation

Field Type Selection Guide:

Data Pattern	Best Field Type	Avoid
Fixed list of options	Single/Multiple Select	Long Text
Yes/No	Checkbox	Single Select with 2 options
Reference to another table	Linked Record	Long Text with copy-paste
Computed from linked records	Rollup	Formula duplicating data
Display from linked record	Lookup	Manual copy
Date with time	Date+Time field	Text field
Money amounts	Currency field	Number field
Sequential counter	Auto Number	Manual number
Calculated from same row	Formula	Manual entry
File/image	Attachment	URL field
User assignment	Collaborator	Single Line Text

2. Schema Audit

Analyze an existing base for anti-patterns:

Anti-Patterns Detected:

• Flat table syndrome — One huge table instead of related tables
• Duplicate data — Same info in multiple tables (not via Lookup)
• Text as enum — Free text where Select field should be used
• Missing relations — Tables that should be linked but aren't
• Formula overload — Complex formulas that should be Rollups
• View chaos — Too many views without naming convention
• Missing primary field — Default "Name" field not meaningful
• Unused fields — Fields with 0% fill rate

Audit via MCP Tools:

1. mcp__claude_ai_Airtable__list_bases → find target base
2. mcp__claude_ai_Airtable__list_tables_for_base → get all tables
3. mcp__claude_ai_Airtable__get_table_schema → detailed field analysis
4. mcp__claude_ai_Airtable__list_records_for_table → sample data for fill rate

3. Schema Documentation

Generate a complete schema doc:

# Base: [Name]

## Tables

### [Table Name] (N records)
| Field | Type | Description | Required |
|-------|------|-------------|----------|
| Name | Single Line | Primary identifier | Yes |
| Status | Single Select | Workflow state (New/Active/Done) | Yes |
| Owner | Linked Record → Users | Assigned person | No |
| Created | Created Time | Auto-generated | Auto |

### Relations
- Users 1→N Projects (via Owner field)
- Projects 1→N Tasks (via Project field)

### Views
- All Projects (Grid, default)
- Active Projects (Grid, filtered: Status = Active)
- By Owner (Kanban, grouped by Owner)
- Timeline (Timeline, by Start/End Date)

4. Migration Planning

When restructuring a base:

1. Map current schema → target schema
2. Identify breaking changes (field renames, type changes, deleted fields)
3. Check for Make.com/Zapier integrations that reference field names
4. Generate step-by-step migration plan:
   • Create new fields first
   • Copy data with formulas or automation
   • Update external integrations
   • Delete old fields last
5. Provide rollback plan

5. Schema Patterns

CRM

Contacts ← Companies (N:1)
Contacts ← Deals (1:N)
Contacts ← Activities (1:N)
Deals ← Pipeline Stages (N:1 via Select)

Project Management

Projects ← Tasks (1:N)
Tasks ← Subtasks (1:N, self-referencing)
Tasks ← Time Entries (1:N)
Projects ← Milestones (1:N)
Users ← Tasks (1:N via Assignee)

Content Pipeline

Content ← Authors (N:1)
Content ← Platforms (N:N via junction)
Content ← Assets (1:N)
Campaigns ← Content (1:N)
Editorial Calendar = Content view by Publish Date

Inventory

Products ← Categories (N:1)
Products ← Suppliers (N:N via junction)
Products ← Stock Movements (1:N)
Warehouses ← Stock Movements (1:N)
Current Stock = Rollup(Stock Movements, SUM(quantity))

Rules

1. Every table must have a meaningful primary field (not default "Name")
2. Use Linked Records instead of duplicating data across tables
3. Prefer Select fields over free text for categorical data
4. Always create at least one filtered View per workflow stage
5. Document the schema before building automations on top
6. Test with 10-20 sample records before bulk import