Your AI team. One flow.

Claude, Codex, Gemini and more — working together on the same project, with the same context. You set the goal; Helmdeck builds the pipeline, assigns the right model to each step, and runs everything in parallel.

Download for macOS Read documentation View prices

Available for macOS · Apple Silicon & Intel

Runs the official CLIs you already use Claude Code Codex CLI Gemini CLI + more
Por qué Helmdeck

Un solo flujo. Tu equipo de IA bajo control.

Orquestación inteligente

Describe el objetivo. Helmdeck arma el pipeline, asigna el modelo correcto a cada paso y ejecuta en paralelo — tú solo apruebas el resultado.

Un mismo flujo, varios agentes

Claude, Codex, Gemini y más — trabajando juntos sobre el mismo proyecto, con el mismo contexto. Cambia de modelo sin perder el rumbo.

Local-first y privado

Tu código nunca sale de tu máquina. Sigues trabajando con los proveedores de IA que ya tienes, y los prompts que genera Helmdeck van directo a ellos.

Smart Pipeline Generator

Describe it. We build the pipeline.

Drop in a prompt, attach a screenshot, mention a file. Helmdeck orchestrates the agents you prefer to analyze your project and lays out the pipeline step by step — each step on the model that fits it best. You stay in control of which steps to add, remove or tweak.

The Board

Every card, a pipeline. The whole flow.

A Kanban where every card is an AI pipeline. The Generate Pipeline column is unique to Helmdeck — describe what you want and the agents you prefer orchestrate the full flow: plan the steps, pick the models, and assemble the pipeline for you to review.

helmdeck · board · 5 columns
LIVE
Backlog
Captured ideas, waiting for triage
Generate Pipeline
Helmdeck drafts sub-cards & prompts from your description
In Progress
Pipelines executing right now
In Review
PRs open, awaiting your approval
Done
Merged, deployed, archived
Worktree Canvas

See your branches, PRs and worktrees as a graph.

Git, demystified. Every worktree, every open PR, every merge — laid out spatially so you can see what's stale, what's blocked, what's ready.

helmdeck · canvas · helmdeck-mac
LIVE
Why Helmdeck

Before Helmdeck vs. after.

You're already paying for the best AI coding tools. Stop running them in separate terminals.

Before

The terminal shuffle

  • ✕ Multiple open terminals, logins, and context windows
  • ✕ Copy-paste prompts between Claude, Codex and Gemini all day
  • ✕ Re-explain your codebase to every new agent
  • ✕ No memory of what ran, when, or which model touched what
  • ✕ Lose your place every time you switch tasks
$ claude $ codex $ gemini $ + más
After

Mission control

  • ✓ One app, all models, shared project context
  • ✓ Pipelines route each step to the best model automatically
  • ✓ Every run logged, replayable, debuggable
  • ✓ Board view of every task across every worktree
  • ✓ Native macOS — fast, quiet, out of the way
C
X
G
+
Use cases

What it solves, day to day.

Three patterns we see every day. Pick a tab, see the pipeline.

Pattern · Refactor

"Migrate this codebase from REST to tRPC."

Helmdeck reads the routes, plans the migration in batches, generates the new code with Codex, and runs Gemini to validate every type contract before committing.

  • ~3 hrs of work, fully automated
  • 32 routes migrated, 0 type errors
  • One PR per logical batch
1
Scan endpoints
Claude
2
Plan batches
Claude
3
Generate tRPC routers
Codex
4
Validate types
Gemini
5
Open PR per batch
Claude
Pattern · Ship a feature

"Add Supabase login with magic links."

Drop a screenshot of the design. Helmdeck wires the auth client, builds the UI, generates the email template, and writes the integration tests — all matching your design system.

  • Design → working feature in one session
  • Tests written alongside the code
  • Onboarding flow drafted by Gemini
1
Read design assets
Claude
2
Scaffold auth client
Codex
3
Build login UI
Codex
4
Magic-link emails
Gemini
5
Integration tests
Claude
Pattern · Debug

"Why is checkout failing in production?"

Paste the Sentry trace. Helmdeck pulls the relevant code, runs three agents in parallel to hypothesize fixes, and ranks them by likely impact before you pick.

  • Three hypotheses, ranked by confidence
  • Each fix opens as a draft PR
  • Tests reproduce the bug before fixing
1
Parse stack trace
Claude
2
Pull related code
Gemini
3
Generate 3 hypotheses
Claude × 3
4
Write failing test
Codex
5
Rank & draft PRs
Claude
Built for the way you ship

Everything you'd expect, and the parts you didn't.

Multi-model routing

Each pipeline step runs on the best model for the job. Helmdeck detects and assigns it for you — and you can change it later if you want.

Worktree-native

Every pipeline runs in its own git worktree. Parallel work, zero merge conflicts.

Approve before commit

Nothing touches main without you. Inline diff, accept, reject, or rerun the step.

Local-first, private

Your code never leaves your machine. Prompts go straight to the providers you trust.

Out of tokens? No problem.

Hit a usage limit mid-run? Helmdeck picks the pipeline back up at the exact step where you stopped — no lost work, no starting over.

Set Sail

Same objective.
Every agent.

Native macOS · Free during Early Access · Bring your Claude, Codex, Gemini accounts. You set the goal — Helmdeck routes the steps and keeps the context locked in.

Download for macOS View prices
macOS 13+ Apple Silicon & Intel