Skip to main content
Ralph Astronaut

Specs drive code.
AI handles the rest.

Write structured specs with OpenSpec. Validate completeness. Let AI implement with full traceability. Also works with Figma, GitHub, Linear, Notion specs.

Get Started →
$ ralph-starter openspec
OpenSpec
? Select a change to build: add-auth (proposal, design, tasks) 87/100
Change: add-auth | Score: 87/100
→ Loop 1/5: Building auth middleware...
→ Loop 2/5: Adding JWT validation + tests...
→ Lint: pass | Build: pass | Tests: 12/12
✓ Done! Cost: $0.42 | PR #42 created
Integrations
OpenSpecFigmaGitHubLinearNotion

Spec Driven Development

One CLI, Five Sources

Pull specs from OpenSpec, Figma, GitHub, Linear, or Notion. Validate. Build. Ship.

ralph-starter
$ralph-starter openspec
OpenSpec
Build from structured specs interactively
? Select a change to build:
> add-auth (proposal, design, tasks, specs/auth) 87/100
Iteration 1/10 | Building auth middleware...
Iteration 2/10 | Writing tests for JWT validation...
Lint: pass | Build: pass | Tests: 12/12 pass
Committed: feat: add JWT auth middleware
PR #42 created: feat: add JWT auth middleware

Browse and select OpenSpec changes interactively. Validates completeness before spending tokens. Learn more →

Ralph Integration

Connect Your Tools

Fetch specs from where your work already lives. OpenSpec directories, GitHub issues, Linear tickets, Notion pages - ralph-starter pulls them in and turns them into working code.

  • ✓OpenSpec structured specs
  • ✓GitHub Issues & PRs
  • ✓Linear tickets
  • ✓Figma designs & Notion pages
$ ralph-starter openspec
? Select a change to build: add-auth 87/100
✓ Change: add-auth (proposal + design + tasks)
✓ Spec validation passed
✓ Starting AI coding loop...
View integrations →
Ralph Coder

AI-Powered Coding Loops

Let AI agents handle the implementation. Ralph runs autonomous coding loops - analyzing, implementing, testing, and committing until complete.

  • ✓Multiple AI agents
  • ✓19 workflow presets
  • ✓Progress tracking
  • ✓Cost monitoring & rate limiting
$ ralph-starter run "build a REST API" --validate
→ Loop 1: Analyzing requirements...
→ Loop 2: Creating endpoints + validation...
→ Loop 3: Adding tests...
✓ Completed in 3 loops | Cost: $0.28
Learn about agents →
Ralph Engineer

Built-in Validation

Every iteration runs through your test suite, linter, and build process. Catch issues early and ensure production-ready code.

  • ✓Auto-run tests
  • ✓Lint checking
  • ✓Build validation
  • ✓Error recovery
$ ralph-starter run "fix the login bug" --validate --shift-left
→ Pre-flight: 1 test failing (captured)
→ Loop 1: Fixing auth token expiry...
→ Tests: 23/23 pass | Lint: clean | Build: pass
✓ All validations passed!
Configure validation →
Ralph Scientist

Git Automation

From commit to PR - ralph-starter handles the git workflow. Each successful iteration becomes a clean commit, ready for review.

  • ✓Auto-commit changes
  • ✓Push to remote
  • ✓Create pull requests
  • ✓Clean commit messages
$ ralph-starter run "add dark mode" --commit --push --pr
→ Committed: "feat: add dark mode toggle"
→ Pushed to origin/feat/dark-mode
→ Creating pull request...
✓ PR #42 created successfully!
Git workflow docs →
New

Auto Mode

Process entire backlogs autonomously. Ralph fetches issues from GitHub or Linear, creates branches, implements each task, runs validation, and opens PRs — all without human intervention.

  • ✓Batch process GitHub issues or Linear tickets
  • ✓Automatic branch creation per task
  • ✓Built-in validation (tests, lint, build)
  • ✓Auto-commit and PR creation
  • ✓Dry-run mode to preview before executing
  • ✓Cost tracking across all tasks
Learn about auto mode →
$ ralph-starter auto --source github --label "ready"
→ Fetching issues labeled "ready"...
→ Found 4 tasks to process
✓ #23 Add user avatar upload
✓ #24 Fix pagination on dashboard
● #25 Add email notifications...
○ #26 Update API docs
✓ 2/4 complete | 2 PRs created | $1.24 total
Get Started

Three steps to
autonomous coding

From installation to your first AI-generated code in under a minute.

Read the full guide →
# 1. Install
$npm install -g ralph-starter
# 2. Write your spec (or use a wizard)
$ralph-starter openspec
✓ Created openspec/changes/add-auth
# 3. Build from your spec
$ralph-starter run --from openspec:add-auth
→ Loop 1: Building auth middleware...→ Loop 2: Adding tests + validation...✓ Done! PR #42 created
Workflow Presets

19 Built-in Presets

Pre-configured workflows for every development scenario. Skip the setup — just pick a preset and go.

⚡Development
featuretdd-red-greenspec-drivenrefactor
🔍Debugging
debugincident-responsecode-archaeology
👁Review
reviewpr-reviewadversarial-review
📄Documentation
docsdocumentation-first
🧪Specialized
api-designmigration-safetyperformance-optimizationscientific-method
$ ralph-starter run --preset tdd-red-green "add user auth"
→ Using preset: tdd-red-green
Max iterations: 50 | Validation: on | Auto-commit: on
Strategy: Write failing test → Implement → Refactor
✓ Test: "should register new user" — FAIL
✓ Implementing auth module...
✓ Test: "should register new user" — PASS
✓ Committed: "feat: add user registration"
View all 19 presets →
Extensible

Agent Skills

Teach your AI agents new tricks. Install community skills or create your own. Skills are auto-matched to your project's tech stack.

  • ✓Global and per-project skills
  • ✓Auto-detect tech stack
  • ✓Browse community skills
  • ✓Create custom skills
Learn about skills →
nextjsfrontend

Next.js App Router patterns

testingtesting

Jest & Vitest best practices

prismabackend

Database schema & queries

tailwindstyling

Utility-first CSS patterns

dockerdevops

Container configuration

securitysecurity

OWASP security patterns

$ ralph-starter skill add vercel-labs/agent-skills
→ Installing skills from vercel-labs/agent-skills
✓ Added 12 skills to ~/.claude/skills/
Use Cases

From spec to shipped

Connect your favorite tools and let AI handle the implementation

OpenSpecOpenSpec

Spec Driven Development

Write structured specs with OpenSpec, validate completeness, then let AI implement with full traceability.

View docs
FigmaFigma

Build React from Figma

Transform Figma designs into production-ready React components with proper styling and responsive layouts.

View docs
GitHubGitHub

Implement GitHub Issues

Turn GitHub issues into working code. ralph-starter reads the issue, understands context, and builds the solution.

View docs
LinearLinear

Ship Linear Tickets

Connect your Linear workspace and let AI handle the implementation of your engineering tickets.

View docs
NotionNotion

Generate from Notion Specs

Write detailed specs in Notion, then let ralph-starter turn them into fully functional features.

View docs

Need a starting point?

Browse our collection of ready-to-use project templates

Browse Templates
Works with all clients
OpenclawOpenclaw
Claude CodeClaude Code
OpenCodeOpenCode
CursorCursor
OpenAI CodexOpenAI Codex
GitHub CopilotGitHub Copilot
Gemini CLIGemini CLI
AmpAmp
Learn More
Use any LLM

Power ralph-starter with your preferred AI provider

Anthropic
AnthropicClaude models
OpenAI
OpenAIGPT-4 & GPT-4o
OpenRouter
OpenRouter100+ models, one API
Configure providers

Integrations

Works With Your Stack

Pull specs from where your work lives. Connect your favorite tools and let Ralph build from them.

OpenSpec

OpenSpec

Spec driven development

Figma

Figma

Design specs, tokens & assets

GitHub

GitHub

Issues, PRs & repository files

Linear

Linear

Project & issue tracking

Notion

Notion

Pages & databases

Contribute on GitHub →
Ralph with integrations