For AI Agent Builders
Your company's data layer for AI agents
Connect any database. Your agent queries it with natural language. Read-only. Scoped API keys. Works with any framework.
Why not just connect directly?
Raw Database MCP
- Agent gets column names only
- Guesses what metrics mean
- Can DROP TABLE
- No usage tracking
- No shared state
AI for Database
- Schema + business context
- Canonical metric definitions
- Read-only guardrailed
- Every query audited
- Cross-agent shared state
How it works
Discover
List sanitized connections, health state, and cached database schemas.
Validate
Run read-only SQL and inspect structured columns, rows, count, and timing.
Draft
Create a paused workflow with an encrypted, destination-bound credential.
Preview
Execute draft queries without contacting the action destination.
Test
Explicitly confirm one live action and inspect sanitized delivery attempts.
Publish
Activate an immutable version and inspect every scheduled production run.
What agents build with this
Autonomous Reporting
Agent pulls metrics on a schedule, posts to Slack
POST /api/v1/chatSmart Customer Support
Agent checks order status before responding to tickets
POST /api/v1/connections/{id}/queryAnomaly Detection
Workflow detects failures, webhooks your agent
POST /api/v1/workflowsInvestor Updates
Agent builds dashboards with 12 months of data
POST /api/v1/dashboardsCross-Database Analysis
Query Postgres + MySQL in one conversation
POST /api/v1/chatData Pipeline Monitoring
Freshness checks + schema drift alerts
POST /api/v1/workflowsLifecycle Automation
Query eligible users on a schedule and send authenticated webhook events
POST /api/v1/workflowsBuilt for agents, not humans in a browser
Smart Connections
When you connect a database, we introspect every table, column, and relationship. Your agent gets a structured schema summary without needing to run discovery queries.
GET /api/v1/connections/{id}/schema
{
"tables": [
{
"name": "orders",
"columns": ["id", "customer_id", "total", "status"],
"relationships": ["customers.id -> orders.customer_id"],
"description": "All customer orders with payment status"
}
]
}Semantic Layer
Define what "revenue" means once. Every agent that queries your database uses the same definition. No more conflicting metrics across tools.
// Metric definition stored in AI for Database
{
"metric": "revenue",
"sql": "SUM(orders.total) WHERE orders.status = 'paid'",
"description": "Total paid order revenue, excluding refunds"
}
// Agent asks: "What is revenue this quarter?"
// AI for Database applies the canonical definition automaticallySaved Queries
Create reusable query templates with parameters. Your agent calls them by name instead of generating SQL from scratch every time.
POST /api/v1/chat
{
"message": "Run saved query: top_customers",
"parameters": { "limit": 10, "period": "last_30_days" }
}
// Returns structured JSON, not raw SQL
{
"data": {
"rows": [...],
"columns": ["customer_name", "total_revenue", "order_count"]
}
}Query Approval
Set rules for which queries run automatically and which need human approval. Critical for production agents that touch sensitive data.
// Approval policy configuration
{
"auto_approve": [
"SELECT on public.*",
"Saved queries tagged 'safe'"
],
"require_approval": [
"Queries touching PII columns",
"Cross-database joins",
"Queries estimated > 10s execution"
]
}Drop-in integration
curl -X POST https://app.aifordatabase.com/api/v1/chat \
-H "Authorization: Bearer afd_your_key" \
-H "Content-Type: application/json" \
-d '{
"message": "Top 10 customers by revenue",
"connectionId": "conn_abc123"
}'Works with everything
Frameworks & Platforms
Databases
Pricing
Enterprise
Custom deployment and SLA
- Self-hosted option
- SSO & audit logs
- Custom rate limits
- Dedicated support