01
Multi-provider support
Built-in templates for OpenAI, Anthropic, Gemini, and OpenRouter. Add any OpenAI-compatible endpoint for internal or experimental infrastructure.
Local-first LLM API workbench
The Postman for LLM APIs.
Stream responses, inspect payloads, manage providers, and generate code snippets — all from one local-first desktop app. No backend, no sign-up.
OpenAI Anthropic Google Gemini OpenRouter Custom endpoints
Features
Everything you need to test LLM APIs.
02
Live streaming
Watch responses arrive in real time. Inspect token flow, partial output, and final payloads in a purpose-built interface.
03
Local history
Every request and response stored locally. Search, replay, compare, and recover the exact payload from any previous run.
04
No backend
Client-only. API keys stay on your machine. Requests go directly to the provider — no hosted relay, no sign-up.
05
Code generation
Export validated requests as cURL, Python, or JavaScript snippets. Move from exploration to implementation in one click.
06
Advanced controls
Attach images, tune temperature and token limits, adjust headers, and work with the parameters engineers actually need.
Product
See it in action.
Workflow
From idea to working request in minutes.
Add a provider
Choose a built-in template or configure your own endpoint, key, custom headers, and model list.
Test the behavior
Iterate on messages, parameters, and streaming to understand what the model does before writing integration code.
Ship the request
Export snippets, keep the run in history, and hand the exact working payload to your app.
Comparison
Why Roshi over alternatives?
| Roshi | cURL / CLI | Postman | Provider playground | |
|---|---|---|---|---|
| Live streaming | ✓ | ✓ | — | ✓ |
| Multi-provider | ✓ | ✓ | ✓ | — |
| Local history | ✓ | — | ✓ | — |
| No account required | ✓ | ✓ | — | — |
| Code generation | ✓ | — | ✓ | — |
| LLM-specific UI | ✓ | — | — | ✓ |
| Keys stay local | ✓ | ✓ | — | — |
| Open source | ✓ | ✓ | — | — |
Security
Your keys never leave your machine.
Roshi is client-only. No sign-up, no hosted relay, no analytics in your provider traffic. Settings and history are stored locally.
What Roshi optimizes for
- latencylow operational drag
- privacylocal key storage only
- clarityraw payload visibility
- portabilitysnippets for real codebases
Pricing
Free and open source.
Roshi is completely free to use. Advanced features will be added in the future.
Current
Free
Everything you need to test and debug LLM APIs.
- All providers supported
- Live streaming
- Local request history
- Code snippet generation
- Vision and image attachments
- No usage limits
Coming soon
Pro
Advanced features for power users and teams.
- Everything in Free
- More features coming soon
FAQ
Common questions.
Does Roshi proxy requests through a backend?
No. Roshi is client-only. Your machine talks directly to the provider. Secrets are stored locally.
Which providers are supported?
OpenAI, Anthropic, Google Gemini, and OpenRouter out of the box. You can also add any OpenAI-compatible endpoint.
Does it support streaming and multi-turn chats?
Yes. Streaming responses are first-class, and the composer supports multi-turn conversations with role-based messages.
Can I generate code from a request?
Yes. Roshi generates cURL, Python, and JavaScript snippets from the request you have validated in the UI.
What are the system requirements?
Roshi runs on macOS 12 (Monterey) and later, including both Apple Silicon and Intel Macs.
Is Roshi available on Windows or Linux?
Not yet. Windows and Linux support are on the roadmap. Star the GitHub repository to follow progress.
What license is Roshi released under?
Roshi is released under the MIT License. You can use, modify, and distribute it freely.
Try Roshi today.
Download the macOS app, connect your providers, and start testing LLM APIs in a purpose-built workspace.