Quick Start¶
Install¶
Install uv if you don't have it, then install Obra:
For voice dictation in Studio, add the voice extra:
Authenticate¶
Obra requires authentication and terms acceptance before first use:
This opens a browser for OAuth. Your code never leaves your machine — only orchestration metadata reaches the Obra service.
Validate Your Environment¶
Obra needs at least one LLM provider CLI installed:
- Claude Code:
curl -fsSL https://claude.ai/install.sh | bash(orbrew install --cask claude-code) - OpenAI Codex:
npm install -g @openai/codex - Gemini CLI:
npm install -g @google/gemini-cli - Ollama: ollama.com (local, no cloud account needed)
Or run obra setup to pick which of your installed providers Obra uses.
Run Your First Session¶
Or use the full guided setup for interactive provider selection with recommended settings:
Launch Workflow Studio¶
For visual workflow authoring, run monitoring, and AI-assisted editing:
If Studio is not ready, diagnose with:
Start the Gateway Directly¶
Use the gateway directly when integrating against the HTTP or WebSocket API:
The API surface is documented in Gateway API.
MCP Integration¶
For OpenClaw, Claude Desktop, Cursor, or any MCP-compatible AI client:
See the Gateway API for the full automation contract.
Other Useful Commands¶
obra doctor # Environment health check
obra briefing quick # Quick context for AI assistants
obra models # Available LLM models
obra help domains # Discover workflow domains
obra config show # View configuration
obra status # Check session status
Desktop App (macOS)¶
A native macOS desktop app is code-complete with guided onboarding, gateway lifecycle management, and a build-verified app bundle. Distribution is pending Apple Developer signing/notarization. See the CLI path above until then.