Create a workflow for `prompts-gpt` before you ever open CI.
This public builder lets you design multi-phase prompts-gpt workflows with real phase settings and export-ready outputs. Build the pipeline, inspect the structure, then run it locally with the package command that fits your execution mode.
Install
npm install -D prompts-gptInitialize
npx prompts-gpt quickstartRun orchestration
npx prompts-gpt orchestrate --mode pipeline --steps pipeline.jsonSequential, parallel, and DAG planning
Model workflows as simple sequences or dependency-aware graphs before you export them.
Phase-level control
Set the tool, model, timeout, retry count, checkpoint behavior, success criteria, and output format for each phase.
Templates to start faster
Load from 10 starter templates, then adapt the pipeline to your repo, team, or task.
Portable exports
Download prompts-gpt pipeline JSON, YAML, Bash, PowerShell, Docker, or GitHub Actions output for local execution and handoff.
Pick the runtime mode that matches the workflow.
The builder is for designing the workflow. The package runtime decides how it executes: provider race, chained steps, or scored evaluation.
parallel
orchestrateRace multiple providers against the same prompt and score the strongest result.
npx prompts-gpt orchestrate --mode parallel -f .prompts-gpt/review.md --providers codex,claude,cursorpipeline
multi-stepChain steps with context passing when research, implementation, and review should remain separated.
npx prompts-gpt orchestrate --mode pipeline --steps pipeline.jsoneval
orchestrateRun once, then force structured scoring so the workflow has a measurable quality gate.
npx prompts-gpt orchestrate --mode eval -f .prompts-gpt/review.md --criteria correctness,risk,clarityInstall the package
Add `prompts-gpt` to the project, then run `npx prompts-gpt quickstart` or `npx prompts-gpt setup` to link the workspace.
Design the workflow
Use the builder below to define the phases, dependencies, prompts, tools, and models that belong in the workflow.
Export and run
Download the pipeline definition or script bundle, then execute it with `prompts-gpt orchestrate`, CI, or a controlled local wrapper.
Design the workflow in public.
Use a starter template or build from scratch. The designer validates the pipeline shape, estimates duration, and generates export output without forcing an account wall.
Phases
1
Est. Duration
5m
Checkpoints
1/1
Rollbacks
0/1
2 warnings
- 1 phase(s) lack success criteria.
- 1 phase(s) lack output format specification.
1 validation error found
Needs fixesPhases (1)
Practical notes before you run the workflow.
The public page is for design and export. Package execution still depends on the tools, providers, and local permissions available in the target workspace.
Why use /orchestration instead of /orchestrate?
The package command is `prompts-gpt orchestrate`, but the public page is a workflow-building surface rather than a single run action. `/orchestration` is the clearer canonical route, and `/orchestrate` can still redirect here.
Can I create a workflow without an account?
Yes. The builder is public and lets you design, validate, and export workflow definitions without signing in. You only need package credentials when you want to sync private prompt packs or run authenticated cloud-backed features.
What can I export from the builder?
You can export prompts-gpt pipeline JSON, generic JSON, YAML, Bash, PowerShell, Docker, and GitHub Actions output. The exports are intended for local execution, CI, review, or team handoff.
Does this match the published prompts-gpt package?
Yes. The page is aligned to the public package workflow documented in the repo README: `quickstart`, `setup`, `sync`, `run`, `sweep`, and `orchestrate` with parallel, pipeline, and eval execution modes.
Next Step
Need prompts before you orchestrate?
Start from the public prompt library or open Prompt Studio first, then come back here when the workflow needs real phase orchestration.