Use Cases
Legible agents running in OpenShell sandboxes can automate a wide range of database workflows — from routine DBA maintenance to complex data engineering pipelines. Each use case below pairs a blueprint (defining the sandbox image, inference profile, and network policy) with a real-world scenario.
These use cases are designed around the three core technologies:
- OpenShell — Provisions isolated, policy-enforced containers where agents execute real commands against your databases
- NemoClaw — Routes inference requests to the right LLM provider and enforces network boundaries so agents can only reach approved endpoints
- Blueprints — Declarative YAML specs that bundle sandbox config, inference profiles, and policies into a repeatable, shareable template
Categories
Database Administration
| Use Case | Description |
|---|---|
| Automated Index Advisor | Analyze query patterns and recommend index changes |
| Backup & Recovery Agent | Schedule, verify, and restore database backups |
| Security Audit Agent | Scan for misconfigurations, stale permissions, and CVEs |
Database Maintenance
| Use Case | Description |
|---|---|
| Schema Migration Agent | Plan, validate, and execute schema changes across environments |
| Performance Tuning Agent | Monitor slow queries, suggest config changes, and apply tuning |
Data Engineering
| Use Case | Description |
|---|---|
| ETL Pipeline Builder | Design and execute extract-transform-load pipelines between sources |
| Data Quality Monitor | Continuously validate data freshness, completeness, and consistency |
Database Creation
| Use Case | Description |
|---|---|
| Database Provisioning Agent | Spin up new database instances with schema, users, and seed data |
| Test Data Generator | Generate realistic synthetic data for development and QA environments |
Data Analysis
| Use Case | Description |
|---|---|
| Exploratory Analysis Agent | Profile datasets, detect anomalies, and surface insights through natural language |
Quick Start
Every use case follows the same pattern:
# 1. Create an agent from the use-case blueprint
legible agent create my-agent --blueprint <blueprint-name>
# 2. Connect to the sandbox
legible agent connect my-agent
# 3. The agent has pre-configured access to your Legible MCP server
# and can query your semantic layer immediately
Or select a blueprint in the Legible UI under Agents → Create Agent → Blueprint, then copy-paste the CLI command shown after creation.