---
title: "Schema"
description: "Canonical provider_token and trigger table definitions"
---
Database adapters expect tables (or collections) with these fields. Column names use snake_case in SQL; Drizzle/Prisma models typically map to camelCase.

## provider_token

Stores OAuth access and refresh tokens per user, provider, and account.

| Column | Type | Notes |
|--------|------|-------|
| `id` | text | Primary key |
| `user_id` | text | Owner user |
| `provider` | text | e.g. `github`, `gmail` |
| `account_email` | text, nullable | Identifies account for multi-account providers |
| `account_id` | text, nullable | Provider-specific account id |
| `access_token` | text | Required |
| `refresh_token` | text, nullable | |
| `token_type` | text | Default `Bearer` |
| `expires_at` | timestamp, nullable | |
| `scope` | text, nullable | Space-separated scopes |
| `created_at` | timestamp | |
| `updated_at` | timestamp | |

### Indexes

```sql
CREATE INDEX provider_token_user_id_idx ON provider_token(user_id);
CREATE INDEX provider_token_provider_idx ON provider_token(provider);
CREATE INDEX provider_token_user_provider_idx ON provider_token(user_id, provider);
CREATE UNIQUE INDEX provider_token_user_provider_account_email_uidx
  ON provider_token(user_id, provider, account_email);
```

The unique index on `(user_id, provider, account_email)` supports multiple accounts per provider. Legacy rows with null `account_email` may coexist until migrated.

### Reference Drizzle schema

Import from the SDK:

```typescript
import { integrateProviderToken } from "integrate-sdk/server";
```

Or copy from `integrate-sdk` source: `src/database/schemas/drizzle.ts`.

## trigger

Stores scheduled tool executions.

| Column | Type | Notes |
|--------|------|-------|
| `id` | text | Primary key |
| `user_id` | text, nullable | Owner; may be repaired via hooks |
| `name` | text, nullable | |
| `description` | text, nullable | |
| `tool_name` | text | MCP tool to execute |
| `tool_arguments` | json/jsonb | Tool arguments |
| `schedule_type` | text | `once` or `cron` |
| `schedule_value` | text | ISO timestamp (once) or cron expression |
| `status` | text | `active`, `paused`, `completed`, `failed` |
| `provider` | text, nullable | OAuth provider for the tool |
| `last_run_at` | timestamp, nullable | |
| `next_run_at` | timestamp, nullable | |
| `run_count` | integer | Default 0 |
| `last_error` | text, nullable | |
| `last_result` | json/jsonb, nullable | |
| `created_at` | timestamp | |
| `updated_at` | timestamp | |

### Indexes

```sql
CREATE INDEX trigger_user_id_idx ON trigger(user_id);
CREATE INDEX trigger_status_idx ON trigger(status);
CREATE INDEX trigger_next_run_at_idx ON trigger(next_run_at);
```

## ProviderTokenData mapping

When the SDK reads tokens, it returns:

```typescript
interface ProviderTokenData {
  accessToken: string;
  refreshToken?: string;
  tokenType: string;
  expiresIn: number;
  expiresAt?: string;
  scopes?: string[];
  email?: string;
  accountId?: string;
}
```

The adapter selects the best row when multiple accounts exist (email/accountId hints from context or OAuth flow).

## Raw callbacks

If you do not use an adapter, implement `getProviderToken`, `setProviderToken`, and `removeProviderToken` manually. See the legacy guide content in the [Database overview](/docs/database) and [Triggers](/docs/database/triggers) pages.