ai-startup-skills
AI 제품을 만드는 스타트업을 위한 7단계 에이전트 스킬: 검증 → 비즈플랜 → 정책 → 로드쇼 → 고객 확보 → 가격 책정 → 등록
아직 감지된 업데이트가 없습니다. GitHub 릴리스가 나오면 여기에 보입니다.
개인 계정의 공개 저장소 소유자는 GitHub로 로그인해 운영자를 확인하고 소개를 직접 관리할 수 있습니다. 관리자 승인은 필요하지 않습니다.
로그인한 뒤 확인 파일을 만들어 이 저장소의 기본 브랜치에 30분 안에 커밋해 주세요. 다른 브랜치나 포크의 파일은 인정하지 않습니다. 파일을 올린 뒤 아래에서 확인하면 관리자 승인 없이 관리할 수 있습니다. 자격을 유지하려면 파일을 남겨 두세요.
로그인이 필요하면 로그인 화면으로 이동합니다. 로그인 후 이곳으로 돌아와 원하는 확인 버튼을 다시 눌러 주세요.
프로젝트 폴더에서 /nomorevibe verify를 실행하면 소유권을 확인하고 소개를 직접 관리할 수 있습니다.
1분 만에 확인하기메이커가 쓴 소개는 위의 한 줄이 전부입니다. 더 자세한 내용은 저장소 README에서 읽을 수 있습니다. 저장소 README ↗
README 전체를 불러오는 중입니다.
README 전문 보기 ↗Paste a London tech event. Get the sponsors and venues most likely to back it, a drafted pitch for each, and a pipeline to chase them.
Live demo · For agents · Run it
Event organisers lose weeks working out who might sponsor them. Backer has already read the sponsor sections of hundreds of London tech events, so it can answer from evidence: who backed similar events, in what role, and where we saw it.
Built at the Grok Bot Community Engineering London Hackathon, 22 Oct 2026. Track 5: Front Desk & Revenue.
Paste an event URL or describe the event. Backer ranks London sponsors and venues by fit (topic, audience, past backing, distance) and shows why each one is on the list. Every reason points at a real past backing with a source link.
Open any lead to get a personal email that names the events that company backed, plus a suggested tier and price. Grok writes the draft and can only cite backings that exist in the map. Send it through Resend, or log a reply when the sponsor answers.
Leads move through contacted, replied, call and signed on a board, and sync to a Notion database. A repeat push reuses the existing page for the same event and company, so a status you changed by hand is kept.
Every sponsor and venue on the map in one table: how many events they backed, the latest one, and every backing with its source link. Open one to draft outreach.
The map as a graph: events, sponsors and the backings between them. Companies that backed the same events sit close together, so you can see who your event shares sponsors with. Outreach and replies are layered on top as memory, stored in Supabase and written to Cognee, which has an "Ask memory" box.
The value is the data. data/london.json is a committed, reproducible map of London tech events, built by scripts/scrape-events.ts from ten sources.
| Events | 316 |
| Sponsors and venues | 547 |
| Backings (sponsor backed event) | 635 |
| Backings with a source URL | 635 of 635 |
Sources: Luma, Devpost, Meetup, Eventbrite, BuilderBase, MLH, dev.events, confs.tech, GDG and Hack Club. Each lives in lib/scrape/sources/ and returns the same RawEvent shape. A source that fails is logged and skipped.
Tech filter. A classifier keeps London tech events and drops the rest. Measured against hand labels:
| Set | Precision | Recall |
|---|---|---|
| Held-out | 1.00 | 0.93 |
| Fresh | 1.00 | 0.95 |
Reproduce with node --env-file-if-exists=.env.local --import ./scripts/register.mjs scripts/eval-tech-filter.ts.
No invented data. A sponsor name only counts if it is literally in the page text or the event's own links. Every backing keeps the URL of the page it was read from, and pitches may only mention backings that are in the data. Details in docs/07-event-scraper.md.
Tell Backer where you are (GPS or a place search) and it uses that for everything.
localStorage. It is never sent to a server to be saved. Without it, Backer assumes central London.Backer is readable by other agents, for example Grok Bot. The MCP server is read-only and uses Streamable HTTP.
| Tool | Does |
|---|---|
search_events | Find scraped events by text, date range, upcoming only, or distance from a point |
find_backers | Rank likely sponsors for an event URL or description, with real past backings as reasons |
get_sponsor | One company by domain: kind, tags, and every event it backed with source links |
MCP client config:
{
"mcpServers": {
"backer": {
"url": "https://grokhack.vercel.app/api/mcp"
}
}
}
REST, same data:
| Route | Query |
|---|---|
GET /api/v1/events | q, lat, lng, radiusKm, from, to, upcoming, limit |
GET /api/v1/backers | event (URL or description), lat, lng, radiusKm, limit |
GET /api/v1/calendar.ics | Events as an iCalendar feed to subscribe to |
curl "https://grokhack.vercel.app/api/v1/backers?event=https://luma.com/your-event&limit=5"
If the server sets BACKER_API_KEY, send it as Authorization: Bearer <key>:
curl -H "Authorization: Bearer $BACKER_API_KEY" \
"https://grokhack.vercel.app/api/v1/events?upcoming=1&lat=51.5236&lng=-0.0786&radiusKm=5"
npm install
cp .env.example .env.local # then fill in the keys you have
npm run dev # http://localhost:3000
Ranking works with no keys at all, because leads come from the committed data/london.json (loaded by lib/seed.ts). Keys switch on the rest:
| Key | Unlocks |
|---|---|
MODEL, XAI_API_KEY | Grok pitch drafts and chat |
SUPABASE_URL, SUPABASE_SERVICE_ROLE_KEY | Shared map and memory storage |
NOTION_TOKEN, NOTION_DATA_SOURCE_ID | Pipeline sync (setup guide) |
RESEND_API_KEY, RESEND_FROM | Sending drafts |
COGNEE_BASE_URL, COGNEE_API_KEY, COGNEE_TENANT_ID | Cognee knowledge graph and "Ask memory" |
Anything unset falls back to the baseline behaviour, so the demo never shows an error.
Refresh the map:
node --env-file-if-exists=.env.local --disable-warning=MODULE_TYPELESS_PACKAGE_JSON \
--import ./scripts/register.mjs scripts/scrape-events.ts
Flags: --past, --future, --limit, --only luma,meetup, --dry-run. XAI_API_KEY adds a Grok pass that finds partners named in event descriptions.
Check types with npx tsc --noEmit. Design and contribution rules are in AGENTS.md; the plan is in docs/PLAN.md.
| App | Next.js 16 (App Router, TypeScript) on Vercel |
| Model | Grok via xAI |
| Data | Supabase Postgres, with the committed data/london.json as fallback |
| Pipeline | Notion API |
| Memory | Cognee |
| Resend |
| Area | Owner | Doc |
|---|---|---|
| Scraper (the map) | Jawad | docs/01-scraper.md |
| Matching and scoring | Avi | docs/02-matching.md |
| Pitch drafts and pricing | Vidit | docs/03-pitch.md |
| Notion pipeline | Joel | docs/04-pipeline.md |
| UI | Dhrishiv | docs/05-ui.md |
| Glue, deploy, demo | Jawad | docs/06-demo.md |
| Event scraper | Avi | docs/07-event-scraper.md |
MIT.