AI-powered CLI agent for Databricks & Snowflake

Your data platform
command line.

An AI-powered terminal agent for Databricks and Snowflake. Debug jobs, prove data quality, analyze spend, and deploy — without leaving your terminal.

$ npm install -g lakecode click to copy copied!
terminal
$ lakecode /debug job 4827
 
Diagnosing job 4827...
Job config loaded — nightly_etl_pipeline
Last 5 runs fetched — 3 failures in 24h
Error logs extracted — OOM on task 3
Query history analyzed — skew on join key
Cluster metrics checked — 98% memory util
 
Root cause: Data skew on customer_id join
causing OOM. Recommend salted join or larger
instance type. Full report saved to ./debug/

What it does

An AI agent that actually understands your Databricks and Snowflake workspace.

AI-powered debugging

  • Diagnose failed jobs end-to-end in seconds
  • LLM analyzes logs, query history, and cluster metrics
  • Get root-cause summaries, not just stack traces

Data quality evidence

  • Generate evidence packs: schema, nulls, freshness, stats, samples
  • AI writes a plain-English health assessment
  • Attach to PRs, tickets, or audit trails

Spend intelligence

  • Top spend breakdown from system billing tables
  • AI detects spikes and explains what changed
  • Audit jobs for missing timeouts, SLA risk, and failure patterns

Deploy & execute

  • Deploy files to your workspace from the terminal
  • Trigger job runs and poll until completion
  • Natural language commands — just describe what you need

How it works

From install to running workflows in under two minutes.

1

Install

Install lakecode with npm. One command, no dependencies to manage.

2

Connect

Sign in with Google OAuth and connect your Databricks or Snowflake workspace. Lakecode automatically profiles your catalogs, schemas, and jobs.

3

Run

Use slash commands or describe what you need in plain English. The AI agent collects context, executes workflows, and delivers actionable results.

Commands

Opinionated workflows for real Databricks and Snowflake engineering tasks.

Debugging & Monitoring

/status job <id> Job config + last 5 runs at a glance
/logs <run-id> Run output + AI error analysis
/debug job <id> Full diagnosis: config, failures, logs, queries, root-cause summary
/queries job <id> SQL query history for the latest run
/queries run <run-id> SQL query history for a specific run

Data Quality

/prove <table> Evidence pack: schema, row counts, nulls, freshness, numeric stats, duplicates, samples + health assessment
/audit jobs Scan for SLA gaps, missing timeouts, recurring failures

Platform Ops

/audit jobs Scan jobs for SLA gaps, missing timeouts, runtime variability, failures; risk-scored report
/cost top Top spend drivers from billing system tables
/cost spike Compare day vs baseline; AI identifies likely drivers
/drift detect Compare deployed resources against bundle definitions

Deployment

/deploy <local> <ws-path> Import a local file into the workspace
/run job <id> Trigger job run + poll until completion
/capture job <id> Capture a job or pipeline into a Databricks Asset Bundle
/uc explain-access Trace why a principal has (or lacks) access to a Unity Catalog object

Why lakecode

AI that understands your workspace

Lakecode profiles your Databricks or Snowflake environment on connect — catalogs, schemas, jobs, warehouses. The AI reasons with real context, not generic prompts.

Evidence, not just answers

Every workflow produces shareable artifacts — evidence packs, debug reports, cost breakdowns. Attach them to PRs, incidents, or audit trails.

Terminal-native, zero context switching

Stop bouncing between Jobs UI, cluster logs, SQL editor, and dashboards. One command replaces five browser tabs.

Security & trust

Runs locally

Lakecode runs on your machine. Your credentials never leave your environment.

Respects your permissions

Lakecode only sees what your platform identity can see. Unity Catalog, Snowflake RBAC — no elevated permissions, no backdoors.

Safe by default

Plan/approval flow for risky operations. Built-in guardrails prevent common high-cost mistakes.

Ed25519 license keys

Authentication uses cryptographically signed license keys. No shared secrets, no tokens stored on our servers.

Get started

Requires Node.js 18+. Two commands and you're in.

terminal
$ npm install -g lakecode
$ lakecode login
 
Opening browser for Google sign-in...
Authenticated as alice@company.com
License key saved
Workspace connected — profiling...
3 catalogs, 12 schemas, 847 tables indexed
 
Ready. Type a command or describe what you need.

FAQ

Is this free?

Yes. Every account gets free daily credits — enough for a few commands per day. Resets every night. No credit card required. For heavier workloads, the Pro plan auto-reloads your balance so you never run out.

Do I need my own API keys?

No. Lakecode includes a built-in AI proxy — just sign in and start running commands. If you prefer, you can bring your own Anthropic or OpenAI keys (BYOK) and use them directly.

What models does it use?

By default, lakecode uses Claude (Anthropic) via our managed proxy. With BYOK, you can use your own Anthropic or OpenAI keys and choose which model to run.

Does lakecode modify my workspace?

Most commands are read-only (debug, prove, cost, audit). Write commands like /deploy and /run job perform actions, but lakecode is explicit about what it's doing and uses plan/approval for risky operations.

Is this just a CLI wrapper?

No. Lakecode is an AI agent with a workflow engine that chains steps, manages workspace context, and produces structured reports. It collects facts via APIs and SQL, then uses AI to interpret, summarize, and recommend.

How does billing work?

Usage-based. Each command consumes credits based on the AI tokens used. Free accounts get daily credits that reset at midnight UTC. Pro accounts add a payment method and auto-reload when the balance gets low. You only pay for what you use.

Simple pricing

Start free. Scale when you need to.

Free

Free daily credits, reset every night. Enough for a few commands per day — debug a job, prove a table, check spend. No credit card required.

Pro

Add a payment method, get auto-reloading credits when your balance is low. Default reload at $20. Usage-based — pay only for what you use.

Documentation

Everything you need to get the most out of lakecode.

Getting started

Install lakecode, sign in with Google, connect your Databricks or Snowflake workspace, and run your first command.

Read →

How it works

The step-based workflow engine: deterministic steps, AI interpretation, context management, and evidence generation.

Read →

Commands reference

Complete reference for all commands: debugging, data quality, platform ops, deployment, and configuration.

Read →

Configuration

Conventions files, default catalogs/schemas, workspace profiles, naming patterns, and team standards.

Read →

Authentication & billing

Google OAuth, license keys, free and Pro tiers, BYOK setup, and managing your account.

Read →

Safety & plan mode

Plan/approval flow, guardrails, output truncation, and how lakecode keeps your workspace safe.

Read →

Evidence packs

What evidence packs contain, how /prove generates them, and how to use them in your workflow.

Read →

Billing & credits

Free tier, auto-reload, cost model, BYOK billing, and managing your balance.

Read →

Start building with lakecode.

Free to try, no credit card required.