Skip to content
Analytics Engineering
Start free
Free and open source

Make your AI coding agent think like an analytics engineer.

Give Cursor, VS Code, or Google Antigravity repository-aware workflows for SQL, dbt, data modeling, metrics, data quality, requirements, and pipelines—without silently connecting to your warehouse.

Install the toolkitView on GitHub

MIT licensed. Local workflows require no AnalyticsEngineering.com account or remote service. A one-click marketplace listing is still pending.

Built by Eric Provencio, an analytics engineer for more than a decade.

AE Toolkit · Local workflow
workspace / analytics
/ae-review-sql models/orders.sql
01Read project conventionscomplete
02Trace query graincomplete
03Review correctness + costactive
04Propose verificationqueued
Paired evaluation

About 40% fewer required checks missed.

The same agent ran each task with and without AE Toolkit. The tasks were kept private until every run had been scored.

Read the method, cost, and limitations
Holdout set A320 runs
40% fewer
required checks missed
Agent alone
21.0%
With AE Toolkit
12.5%
Holdout set B287 runs
38% fewer
required checks missed
Agent alone
18.2%
With AE Toolkit
11.2%

Runs used Cursor Composer 2.5 on synthetic repositories. Set A was statistically clear; set B pointed in the same direction with a smaller, inconclusive margin.

Featured workflows

Start with the work analytics engineers do every day.

These six entry points are only the beginning. The public repository currently ships 25 agent skills and 18 command aliases spanning the analytics development lifecycle.

  • /ae-review-sqlAvailable locally
    Workflow 01

    Review SQL

    Check correctness, grain, joins, performance, and production risk.

  • /ae-profile-dataAvailable locally
    Workflow 02

    Profile data

    Compute and interpret a local CSV, JSON, or JSONL profile.

  • /ae-review-modelAvailable locally
    Workflow 03

    Review a data model

    Assess or propose a model from business and source evidence.

  • /ae-dbt-healthAvailable locally
    Workflow 04

    Check dbt health

    Inspect a real dbt project without blindly executing dbt.

  • /ae-requirementsAvailable locally
    Workflow 05

    Clarify requirements

    Turn an ambiguous request into definitions and acceptance criteria.

  • /ae-explainAvailable locally
    Workflow 06

    Explain data code

    Explain selected SQL, Python, dbt, YAML, or pipeline code.

Explore every workflow on GitHub
Why it is different

Grounded in the repository, not generic advice.

Learns the house style

It inspects naming, keys, tests, documentation, SQL, warehouse, semantic-layer, and BI conventions before proposing work.

Asks fewer, better questions

It infers what the repository can prove and asks only for correctness-critical business facts such as grain or metric meaning.

Keeps risky actions visible

It plans bounded validation, protects production-like targets, and does not silently upload workspace content or mutate warehouse data.

Product boundary

Analysis first, with visible control.

  • It does not silently connect to a warehouse or mutate production.
  • It does not replace dbt, your warehouse, source control, or a BI tool.
  • Connected course grading and remote services are not available today.
  • The local plugin contains zero anonymous telemetry.
Common questions

What to know before you install it.

What is the Analytics Engineering Toolkit?
AE Toolkit is a free, open-source plugin and collection of Agent Skills that gives AI coding agents repository-aware workflows for SQL, dbt, data modeling, metrics, data quality, and analytics pipelines.
Which AI coding agents does AE Toolkit support?
Cursor gets the full plugin and enforced safety hooks. VS Code and Google Antigravity get generated workspace integrations. Claude Code, Codex, Gemini CLI, OpenCode, and other Agent Skills hosts can install the skills.
Does AE Toolkit connect to my data warehouse?
Not silently. The toolkit runs locally, has no telemetry, and asks for approval before a query, warehouse-executing dbt command, or dbt MCP call can run.
Has AE Toolkit been tested?
Yes. In paired tests on two sets of previously unseen analytics-engineering tasks, the same agent missed 40% and 38% fewer required checks with AE Toolkit installed. The public evaluation documents the method, run counts, cost, and known weakness.

Marketplace updates

Want one-click installation later?

The toolkit is already public and installable from GitHub. Join the list for the Cursor Marketplace release and major toolkit updates.

No spam; unsubscribe anytime.