AGENTS
Teachable AI that speaks your language.
◇
Teach Your Domain
Every business has its own vocabulary. Teach agents your terminology, rules, and context with 16 teachable types.
teachable notebook
+ Add teachable
term"MRR" = Monthly Recurring Revenue - sum of all active subscription values.
glossaryrevenue = SUM(sales_amount) - Maps business term to SQL expression.
alias"the big table" = orders - User shorthand.
explain"Churn" = Customer who cancels subscription. Use churned_at IS NOT NULL.
analogySubscription = Think of a subscription like a gym membership.
16 Teachable Types Across 4 Categories
DOMAIN (5 types)
term · glossary · alias · explain · analogy
Define vocabulary, map terms to SQL, create shortcuts
RULES (4 types)
hint · guardrail · styleGuide · quirk
Behavioral rules, safety constraints, SQL style, data edge cases
EXAMPLES (3 types)
example · workflow · clarification
Q&A pairs for few-shot, multi-step processes, when to ask clarity
IDENTITY (3 types)
persona · preference · correction
AI personality, output format, fixes
◇
Shortcuts
Configure quick commands that appear in the chat interface. Control when they're visible and what they do.
command palette
/ Top customers by revenue
/ Revenue trend this quarter
/ Active subscriptions by plan
/ Churn rate by month
Visibility:
- •Custom prefix, label, and description
- •Control visibility (before/after interaction, always)
- •Immediate execution option
- •Drag-to-reorder
◇
Knowledge Hub
Auto-generate business context from your schema using AI. The agent learns table relationships and business logic.
generated report
## Business Overview
This database tracks customer subscriptions and billing...
## Key Entities
- customers: Core customer records with contact info and status
- subscriptions: Active and historical subscription records
- invoices: Billing records linked to subscriptions
## Relationships
- customers → subscriptions (1:many)
- subscriptions → invoices (1:many)
## Business Rules
- Customers can have multiple active subscriptions
- Invoices are generated monthly on subscription anniversary
- •Uses LLM to analyze your schema and generate context
- •Detects when grounding config changes (staleness indicator)
- •One-click regeneration
- •Toggle on/off per agent
◇
Grounding Control
Control exactly what the agent can see. Filter tables, hide sensitive columns, configure introspection depth.
grounding config
TABLES
Filter mode:
✓
customers"Core customer records"✓
orders"Purchase transactions"✓
products"Product catalog"internal_logs(hidden)
admin_settings(hidden)
RELATIONSHIPS
Forward depth (child → parent):3 ▼
Backward depth (parent → child):1 ▼
METADATA
✓
IndexesQuery optimization hints✓
ConstraintsCHECK, UNIQUE, NOT NULL, DEFAULT✓
Row countsSize classification (small/medium/large)✓
Column statsMin/max values, null fractions✓
Column valuesEnum-like detection and caching- •Filter tables by name list or regex pattern
- •Control FK traversal depth (1-10 or infinite)
- •Configure metadata extraction (indexes, constraints, stats)
- •Add custom descriptions per table
◇
Model Selection
Choose your preferred AI provider and model. Bring your own API keys.
providers
OpenAI
Anthropic
Groq
Gemini
Ollama
Azure
Direct API calls. No proxy. Your keys.