翻訳待ち：Show HN: Backprompter – create, test, and deploy agents without a back end

AI サービスが一時的に利用できないため、復旧後に翻訳を補完します。

Backprompter Build, test, and ship AI agents — no backend code. A complete workspace to create AI agents from system prompts, chat with and evaluate them, then deploy a polished chat UI for your users — with API access, RAG, voice, and Stripe billing built in. ✍️ BuildSystem prompts, combo-agents, and a meta-router that picks the right agent per message. 🧪 TestMock users, simulated conversations, scored evaluations, and full version history. 🚀 ShipOne-click chat UI, REST API & SDK, file/RAG, voice chat, and direct Stripe payments. 🔒 Zero data retention — your prompts & outputs are never stored by the model providers You're signed in as . From prompt to production in three steps No servers to configure, no boilerplate to write. 1 · CreateSpin up an app and define your agent with a plain-language system prompt. 2 · TestChat with it, simulate real users, and score responses to refine its behavior. 3 · DeployPublish a hosted chat UI or call the API & SDK — nothing to maintain. Ship it your way The same agents, deployed however fits your stack — from zero code to full control. No code Hosted chat UI Publish a polished chat interface for your end users — no frontend or backend to build. Pick your sign-in methods and share the URL. $ no code required ✓ live at your-app.web.app/chat How deploying works → Your frontend Custom UI with the SDK Build your own branded interface while Backprompter handles authentication, agent calls, and per-user data — no backend to run. Backprompter.setAppKey('your-app-key'); await Backprompter.signInAnonymously(); const { response } = await Backprompter.run({ agentId: 'support-bot', userPrompt: message, }); Frontend SDK guide → Your backend Call agents from your server Use a Backprompter API key from your own backend. You handle user auth however you like — Backprompter never sees your users' credentials. const bp = new BackprompterServer({ apiKey: process.env.BACKPROMPTER_API_KEY, tenantId: process.env.BACKPROMPTER_APP_KEY, }); const { response } = await bp.run({ agentId: 'support-bot', userPrompt: ticket, }); Backend SDK guide → Prefer to self-host? Run Backprompter locally with your own models or other providers — contact us for a quote. Loading your workspace... Voice idle Current Comparing Prefetch default agent (run default agent in parallel with routing) Routable agents Deploy a chat UI End-user login methods (visible to end users when the UI is deployed) End-user login methods Anonymous Secret code Email/password Google Allow end-users to create their own accounts (sign-up) Host it on your own domain Download the chat frontend, pre-configured for this app. Modify it freely and deploy to any static host or your own domain. Embed as a chat box Add the chat to an existing website with an iframe: Secret keys (one-time codes) New secret (store now — will not be shown again) Actions Key Created Warning: If you uncheck Secret code and deploy, end-user secret-code access will be disabled. Charge your end-users (Stripe) Connect your own Stripe account to require payment before end-users can use this app. Payments go directly to you — Backprompter never touches the money and takes no fee. ✓ Stripe connected Price (from your Stripe account) Require payment to use this app Description User-Facing App Name (optional) Logo URL (optional) Background Image URL (optional) Bring Your Own Keys (BYOK) OpenRouter API Key OpenAI API Key Leave blank to use default keys. Keys are encrypted and never displayed after saving. Summary System Prompt For combo-agents: Describe how to coordinate sub-agents (e.g., "1. Call Agent1... 2. If result is X, call Agent2...") Routable through meta-router Include user profile context Enable web search (OpenRouter models only) Enable web fetch (auto-fetch URLs found in user prompt) Model Any OpenRouter model ID. Leave blank to use the app default. Temperature (0-1) Max Tokens Artifact Type Artifacts appear in a side pane instead of the chat Tools (read-only GET tools this agent may call) Define tools once at the app level, then pick the ones this agent should use. RAG Augment agent context with documents from your app System Prompt Select a version to view the system prompt Versions Summary Detailed Profile Agent Version Mock User Response Feedback Rating Builder Timestamp Mock User Updated Profile Profile Changes Which agent did you evaluate? Evaluator Agent Evaluator Response Human Feedback (max 100 words) 0 / 100 words Rating (0-10) Suggested System Prompt Update tags back here. -->