Customer Engagement Platform for Email, Push Notifications, & SMS
The world's leader for mobile push notifications, web push, SMS, email and in-app messaging. Trusted by 2 million+ businesses to send 12 billion+ messages per…
翻译准备中
尚未检测到更新。GitHub 发布版本后会显示在这里。
个人 GitHub 账号下公开仓库的所有者可以登录并验证所有权,自行管理介绍,无需管理员批准。
登录并生成验证文件,在 30 分钟内将其提交到此仓库的默认分支。其他分支或分叉仓库中的文件无效。然后在下方检查文件,即可无需管理员批准直接管理项目。请保留文件以维持管理权限。
如需登录,将前往登录页面。登录后请回到此处,再次点击所选的验证按钮。
在项目文件夹中运行 /nomorevibe verify,即可验证所有权并自行管理介绍。
一分钟完成验证Drop a name or any social URL. Get a personalised video in 60 seconds.
翻译准备中
取自仓库 README
正在加载完整 README。 以下为已保存的摘录。
The conversational SDR that looks and sounds like you.
nuncio is an AI sales agent for founders and small B2B teams pursuing high-value accounts, partnerships, investors, and other conversations where one thoughtful first message can change the relationship.
We start with a personalised video because it is the best hook in the world. The real product is a live, conversational AI avatar of the sender — trained on their face, voice, and playbook — that can talk to prospects, answer questions, and book meetings.
No templates. No mail merge. A real agent for the conversations that matter most.查看完整 README ↗The conversational SDR that looks and sounds like you.
nuncio is an AI sales agent for founders and small B2B teams pursuing high-value accounts, partnerships, investors, and other conversations where one thoughtful first message can change the relationship.
We start with a personalised video because it is the best hook in the world. The real product is a live, conversational AI avatar of the sender — trained on their face, voice, and playbook — that can talk to prospects, answer questions, and book meetings.
No templates. No mail merge. A real agent for the conversations that matter most.
/v/[id]) with sharing, captions, and optional translation.Total time from input to video: ~5 minutes.
nuncio uses Band as the coordination layer. Four specialized agents work together in a shared room:
User joins Band Room → posts prospect URL + brief
│
▼
┌───────────────────┐
│ Researcher Agent │ ← Enriches profile via TinyFish
└────────┬────────┘
│ (structured message)
▼
┌───────────────────┐
│ Copywriter Agent │ ← Generates angles + scripts via Claude/Featherless
└────────┬────────┘
│ (drafts)
▼
┌───────────────────┐
│ QA Agent │ ← Validates word count, brand safety, pronunciation
└────────┬────────┘
│ (approved/revision request)
▼
┌───────────────────┐
│ Producer Agent │ ← Renders audio (ElevenLabs) + video (HeyGen)
└────────┬────────┘
│
▼
Video link + landing page
Users see agent reasoning, approve angles, and track progress in real-time. Human approval required before video renders.
| Layer | Technology |
|---|---|
| Coordination | Band — multi-agent room |
| Enrichment | TinyFish Fetch + Search API |
| Intelligence | Anthropic Claude or Featherless AI |
| Audio | ElevenLabs — soundscape, cinematic entrance |
| Video | HeyGen Video Agent API, Avatar V |
| Speech | Speechmatics — captions, transcription |
| Storage | Turso (SQLite) + file fallback |
| Frontend | Next.js 16, TypeScript, Tailwind CSS, Motion |
| Deployment | Docker, Vultr + Coolify |
git clone https://github.com/udirobert/nuncio
cd nuncio
cp .env.example .env.local
# Add your API keys to .env.local
pnpm install
pnpm dev
Visit http://localhost:3000?demo=true to see the full flow with cached data (no API keys needed).
# Required
TINYFISH_API_KEY=
HEYGEN_API_KEY=
HEYGEN_AVATAR_ID=
HEYGEN_VOICE_ID=
# Intelligence (at least one)
ANTHROPIC_API_KEY=
FEATHERLESS_API_KEY=
# Optional
ELEVENLABS_API_KEY=
SPEECHMATICS_API_KEY=
TURSO_DATABASE_URL=
NEXT_PUBLIC_APP_URL=http://localhost:3000
The video is the hook. The long-term goal is a live, conversational AI avatar that continues the conversation on the sender's behalf.
Recipient→sender signup (?ref= links) is the loop — run the invite-flow teardown + K-factor worksheet on it, then a 48-hour test on share-CTA copy. Ladder check: core flow proven; find the first unproven rung (spreads → hops → pays) and stop work above it.
Accepted into the Keiki × Maritime programme (Oct 20 – Dec 1, 2026): build an always-on business agent, reach 1,000 users in six weeks. Transparent user tracking starts at zero: docs/KEIKI_MARITIME.md. The STRATEGY scoreboard (docs/STRATEGY.md) tracks conversion and engagement per artifact; the Keiki × Maritime board tracks cumulative users against the programme goal.
/v/[id]) is a marketing surface with "Make your own" CTA?demo=true runs the full pipeline with cached data for presentationsnuncio/
├── src/
│ ├── app/
│ │ ├── page.tsx # Main pipeline UI
│ │ ├── playbook/page.tsx # Worked examples
│ │ ├── v/[id]/page.tsx # Video landing page
│ │ └── api/
│ │ ├── enrich/route.ts
│ │ ├── script/route.ts
│ │ ├── video/route.ts
│ │ └── ...
│ ├── components/
│ │ ├── url-form.tsx
│ │ ├── angle-picker.tsx
│ │ ├── script-review.tsx
│ │ └── ...
│ └── lib/
│ ├── tinyfish.ts # Enrichment
│ ├── claude.ts # Script generation
│ ├── heygen.ts # Video rendering
│ ├── elevenlabs.ts # Audio generation
│ └── pipeline.ts # State machine
├── docs/
│ ├── ARCHITECTURE.md
│ ├── BAND_INTEGRATION.md
│ ├── DEPLOY.md
│ └── ...
├── Dockerfile
└── .env.example
See docs/DEPLOY.md for deployment guide (Vultr + Coolify, Docker, or direct VPS).
Thin MCP/HTTP wrapper for chatgpt.com/plugins. Description in user words. Free research_and_draft first (quick-tier research + channel draft in-chat — discovery + mid-conversation recommendation). Render / live link second via an existing account or informational plans page on the product site — not in-plugin checkout (OpenAI allows plugin commerce for physical goods only).
| Tool | Price | Endpoint |
|---|---|---|
research_and_draft | Free | POST /api/mcp/research-and-draft |
| Service discovery | — | GET /api/mcp/tools |
| Thin MCP stub | — | POST /api/mcp (tools/list, tools/call) |
# Demo path (no live research) — labelled sample draft
curl -X POST http://localhost:3000/api/mcp/research-and-draft \
-H 'content-type: application/json' \
-d '{"demo":true,"url":"https://example.com"}'
Playbook: docs/CHATGPT_PLUGIN_PLAYBOOK.md · Connect: docs/CONNECT.md · Eval: docs/EVAL.md · Starters: docs/STARTER_PROMPTS.md · Scoreboard: usage-only funnel in docs/CHATGPT_PLUGIN_PLAYBOOK.md (usage-first; don’t chase ARPU yet)
Neatlogs tracing is shipped, opt-in on NEATLOGS_API_KEY — see docs/NEATHACK.md for the trace map, detections, and demo script.
band.pipeline_run → research/script/review/render/media spans) + Hermes loop (hermes.enqueue, hermes.prospect_run, hermes.reply_classify, hermes.earn_checkout) + agent.lite_run and mcp.research_and_draft; per-provider llm.* spans inside the fallback chainsrc/lib/neatlogs.ts (traced() no-ops without the key; errors are recorded then re-thrown)ssh nuncio-vultr 'bash -s' < scripts/neathack-verify.sh (wiring + span-name coverage + doctor --probe finalization + one live trace)AgentCribs "startup's first hire": nuncio as OpenClaw 2.0 multiplayer SDR. Submit by Sep 29, 11:59pm PT to the AI Worth Using Agent Index (MIT, usage reporting); leaderboard snapshot Sep 30.
/v/[id] + /live/[id] share links are the distributionMIT