FREE TOOL · AGENTIC CODING & DATA ENGINEERING

vibeCodeCLI

Coding agents are non-deterministic. Run the same prompt twice, get two different results. vibeCodeCLI races parallel agent fleets, remembers your past decisions, and checks outcomes before running anything irreversible, now across your codebase and your entire data stack. Free. Bring your own API keys.

$ curl -fsSL vibecodecli.com/install.sh | sh copy $ npm install -g vibecodecli copy
vibecodecli.com / product intro
The platform

Your code and your data stack in. Shipped work out.

One CLI between your repo, your databases, and production. Fleets race in the middle, memory recalls what worked, and a safety gate checks outcomes before anything irreversible.

«DEVICE» YOUR MACHINE · «COMPONENT» vibeCodeCLI 23 PROVIDERS · BRING YOUR OWN KEYS OUTBOUND HTTPS ONLY 15 CONNECTORS, ONE INTERFACE SECRETS MEMORY GATE GROUND 01/ «ground» introspects real schema & dbt lineage first 02/ «gate» EXPLAIN cost preview · outcome check 03/ «memory» recalls schema quirks & past fixes 04/ «secrets» credentials.local · never leave this machine YOU task in plain English TASK «INTROSPECT» «DATABASE» PostgreSQL · MySQL Mongo «QUERY» «WAREHOUSE» Snowflake · BigQuery Redshift «EXPLAIN FIRST» «ENGINE» Databricks ClickHouse · DuckDB «LINEAGE» «STORAGE · DBT» S3 dbt project & lineage «MCP» «PLATFORM» Bonacci Studio run · monitor pipelines
PAR [3 isolated git worktrees] you vibeCodeCLI agent-1 agent-2 agent-3 ThinkingMemory 01/ run --agents 3 "add rate limiting" 02/ recall(task context, past decisions) 03/ conventions · prior outcomes 04/ «spawn» build · test build · test build · test 05/ ✗ tests fail 06/ ✓ 41 tests pass 07/ ✓ pass · slower 08/ outcome gate · simulate irreversible actions 09/ remember(winner · outcome) 10/ winning diff «agent-3» → review & merge
Your stack
▸A task in plain English
❖Git repository
⛁Databases & warehouse
▤dbt project
>_vibeCodeCLI
fleetN agents race in isolated worktrees
memoryThinkingMemory recalls past decisions
groundreal schema and lineage before queries
gateoutcome check before irreversible actions
23 providers · bring your own keys · runs in your terminal
What ships
✓The winning diff, merged
✓Grounded SQL & models
→Pipelines in Bonacci Studio
Why it is different

Three problems with coding agents, solved

01 / parallel fleet

Do not bet on one agent

N agents race the same task in isolated git worktrees. You review the diffs and merge the winner. Git is your undo.

02 / persistent memory

It stops forgetting

A ThinkingMemory backbone recalls past decisions and outcomes into every turn, so the agent builds on what it already knows about your codebase.

03 / outcome-aware

It can walk away safely

Before any irreversible action a forward check simulates the outcome. Commands that would destroy data are rejected, even in full-auto mode.

Data Engineering Agent

The Data Engineering Agent that runs your data stack

Point vibeCodeCLI at any database, warehouse, or dbt project, or hand it off to Bonacci Studio, and the same safety model applies.

grounded

Reads your schema first

Connects to your databases, warehouse, and dbt project, introspects the real schema and lineage before it writes a single query.

15 connectors

Runs across your stack

PostgreSQL, Snowflake, BigQuery, Redshift, Databricks, ClickHouse, DuckDB, MongoDB, S3 and more, behind one interface.

credential-local

Your passwords never leave

The runner stays on your machine, outbound HTTP only. EXPLAIN previews cost before anything executes.

vibecodecli → studio

Ships pipelines to Bonacci Studio

Connect with one token and it designs, runs, and monitors pipelines in Bonacci Studio, from a database pull to a Spark job, without leaving the terminal.

vibecodecli.com / data engineering
23
model providers supported
15
database & data stack connectors
N×
parallel agents per task
$0
no subscription, free
vibecodecli → studio

When the pipeline outgrows the terminal

Connect once with a token and keep working from the CLI while Bonacci Studio runs, schedules, and monitors the pipelines you ship. Same agent, a production home for the work.

Get started

Stop gambling on one agent

Free. Bring your own API keys and pay your provider directly. No subscription, no lock-in. Works on your code and your entire data stack.

~/your-project
# install, free and bring your own keys
$ curl -fsSL vibecodecli.com/install.sh | sh
  ✓ vibecodecli ready

# race a fleet on one task
$ vibecodecli run --agents 4 "add rate limiting"
  agent-1  building...
  agent-2  building...
  agent-3  ✓ tests pass
  agent-4  building...
  → review the diffs, merge the winner

# or point it at your data stack
$ vibecodecli connect snowflake
  ✓ schema and dbt lineage grounded
  → credentials stay on this machine

# or hand off to Bonacci Studio
$ vibecodecli connect studio bcs_pat_...
  ✓ studio connected
  → design, run, and monitor pipelines from here