YC Outreach

Pick any Y Combinator batch, get every company's founders and their likely email addresses, and write a personalised cold email to each one from a single template. Free by default (no API keys, no sign-up); optional verified emails with your own Apify token.

python3 serve.py        # open http://localhost:8765

Python 3.9+, standard library only. Nothing to install.

1. Pick a batch (Summer 2005 through the latest) from YC's public directory.

2. Founders load 20 companies at a time (about 5–10 s per 20); click Load more for the next 20. For each company the server reads its ycombinator.com page for the founders' names, titles, LinkedIn and X, and for how many roles that company has open on YC (title and link for each). Loaded companies are kept in your browser, so reopening a batch is instant. Check hiring to hide companies with no open roles.

3. Emails are filled in for each founder, best source first:

   Label	Source	Reliability
   verified	Apify lookup with your token (optional)	Checked by Apify
   unverified	Apify lookup that couldn't confirm the mailbox	Usually right
   on site	Email published on the company's homepage or /contact page	Real address, may be generic
   guess	first@domain, then first.last@, flast@, firstlast@	First guess right ~80% of the time

   The reliability figure comes from a check against 151 Apify-verified founder emails: the first guess matched 80%, and the right address was somewhere in the guess list 93% of the time.

4. Write once, send many. Fill in your name and links, edit the subject and body, and every draft updates live. Copy the text or open it in your mail app. Mark companies as sent to hide them.

Your details, template, "sent" marks and loaded batches are stored in your browser (localStorage). There's no database and no account.

Import the repo on Vercel. No build step, no environment variables. index.html is served as a static page and api/yc.py runs as a Python serverless function (vercel.json gives it 60 s).

All endpoints are GET and return JSON. Errors return {"error": "..."} with status 400 (bad input) or 502 (YC unreachable).

[{"batch": "Fall 2026", "count": 110}, {"batch": "Summer 2026", "count": 231}]

Newest first. Includes "Unspecified", which the UI hides.

[{"name": "Indemni", "slug": "indemni", "batch": "Winter 2024", "website": "http://www.indemni.com",
  "one_liner": "Cargo Theft and Fraud Prevention Platform", "industry": "B2B -> Supply Chain and Logistics",
  "team_size": 7, "launched_at": 1708029636}]

batch must look like Winter 2024 (season + year).

At most 10 slugs per call (^[a-z0-9-]+$). The UI loads 20 companies per click as two parallel calls. About 5–9 s per call; each company's website check is cut off after 4 s (SITE_DEADLINE) so one slow site can't stall the batch.

[{"slug": "indemni", "website": "http://www.indemni.com", "domain": "indemni.com",
  "linkedin": "https://www.linkedin.com/company/...", "twitter": "", "site_emails": [],
  "job_count": 1, "jobs": [{"title": "Founding Engineer", "url": "https://www.ycombinator.com/companies/indemni/jobs/..."}],
  "founders": [{"name": "Omar Draz", "title": "Founder", "linkedin": "https://linkedin.com/in/odraz",
                "twitter": "https://twitter.com/oamdraz", "emails_found": [],
                "email_guesses": ["omar@indemni.com", "omar.draz@indemni.com", "odraz@indemni.com", "omardraz@indemni.com"]}]}]

A company whose YC page fails to load comes back as {"slug": "...", "error": "..."}. Guesses are empty when the domain doesn't resolve.

• Batches and companies: YC's public company search (Algolia). The read-only search key is read from ycombinator.com/companies at runtime, so no key is stored here.
• Founders and open roles: the data-page JSON embedded in each ycombinator.com/companies/<slug> page (props.company and props.jobPostings). Incomplete postings are skipped. Browsers can't fetch these cross-origin, which is why this part runs on a server.
• Verified emails (optional): the Apify actor snipercoder/email-finder-by-name-and-domain, called from the browser with the visitor's own token, so the token never reaches the server. Results are matched on the Name and Domain fields of each dataset item. Each run covers up to 50 founders and is capped with maxTotalChargeUsd.

For bulk exports. Output files are gitignored.

python3 yc_scraper.py --batches "Winter 2024" "Summer 2024" --out yc_founders   # -> yc_founders.json + .csv
python3 apify_enrich.py --input yc_founders.json --out yc_founders_enriched      # needs APIFY_TOKEN

yc_scraper.py flags: --batches (default "Summer 2026" "Fall 2026"), --out, --workers (default 12), --limit (first N companies, for testing). It checks six contact pages per site and retries, so it's slower but more thorough than the web API.

apify_enrich.py reads APIFY_TOKEN from the environment or a .env file (gitignored). Flags: --input, --out, --limit, --chunk (founders per run, default 50), --max-charge (USD cap per run, default 1).

• Standard library only, on purpose. Don't add dependencies or a build step.
• api/yc.py and yc_scraper.py share logic but are separate on purpose: the function has to finish inside a serverless timeout, so it makes one attempt per fetch, checks 2 pages, and gives up on a website after 4 s; the CLI retries and checks 6.
• The function only accepts a batch name or slugs. Websites always come from YC's data, never from the request, so it can't be used to fetch arbitrary URLs. Keep it that way.
• Everything from YC is untrusted text. index.html escapes it (esc()) and only links http(s) URLs (url()).
• To test locally: python3 serve.py, then curl 'localhost:8765/api/yc?action=founders&slugs=reddit'.

Write to people one at a time, keep it short, and take "no" for an answer.

MIT