Skip to main content

Using Atlas with AI Agents

AI agents are great at writing code, but generating database migrations is a different challenge. As schemas grow, ensuring migrations are deterministic, safe, and policy-compliant becomes critical.

Atlas lets AI agents focus on editing the schema while it handles the infrastructure:

  1. Migration Generation — correct, safe, and deterministic migrations.
  2. Migration Linting — semantic validation and best-practice enforcement.
  3. Policy Enforcement — organizational rules for allowed changes.
  4. Schema Testing — AI writes logic (functions, views, queries) and tests; Atlas executes and reports failures.
  5. Data Migration Testing — seed data, run tests, detect errors, and let the AI fix them.

Agent Skills​

Agent Skills are an open standard for packaging domain expertise for AI agents. The agent runs the Atlas CLI instead of learning it from scratch. Atlas provides these skills:

  • atlas covers schema migrations, linting, and testing, Data Scripts for backfills and reports, and Atlas Cloud operations: database status, failed deployments, and what is waiting for deployment.
  • atlas-onboard takes a project from a first scan to production, one verified stage at a time.
  • atlas-postgres and atlas-mysql add what is specific to PostgreSQL, and to MySQL and MariaDB. The agent loads the matching one when the project uses that database (Database Skills).
  • atlas-operator covers the Atlas Kubernetes Operator: installing it, writing its resources, deploying through GitOps, and troubleshooting (Kubernetes Operator Skill).

Claude Code installs all of them as a plugin. Other agents install them with npx skills:

# Claude Code
claude plugin marketplace add ariga/atlas
claude plugin install atlas@ariga

# Codex, Cursor, GitHub Copilot, and other agents
npx skills add ariga/atlas

To help agents produce better results, provide structured schema context. See the Database Schema as Context guide for how to organize schema files and the benchmark showing 10/10 query accuracy with the right structure.

A machine-readable docs index with usage terms is available at atlasgo.io/llms.txt.

Agent Skills

Install the Atlas skills for Claude Code, Codex, Cursor, and Copilot: migrations, Data Scripts, Atlas Cloud status, and guided onboarding.

Database Schema as Context

Organize schema files for maximum AI agent accuracy.

Setup Instructions​

Configure your AI agent to work with Atlas using project-level instruction files:

GitHub Copilot

Configure GitHub Copilot with Atlas-specific instructions.

Cursor

Set up Cursor with Atlas-specific rules.

Claude Code

Set up Claude Code with Atlas-specific instructions.

OpenAI Codex

Set up OpenAI Codex with Atlas-specific instructions.

Example Workflow​

The instructions above teach the AI agent to follow this workflow when generating migrations:

1. Edit the schema​

Editing the schema

2. Generate and validate migrations​

The agent runs atlas migrate diff to generate the migration, then atlas migrate lint to validate it.

Generating and linting migrations

3. Apply the migration​

The agent applies with atlas migrate apply, starting with a dry-run.

Applying the migration