Avinashistoocool / Agnostos-v2

# Agnostos: AI research internship outreach
**Agnostos** is a free, open-source tool that helps students land research internships. It finds professors whose recent papers match your resume, drafts one honest cold email per professor, sends them slowly from your own inbox during each professor's working hours, and drafts replies to the answers.
> Hosted web version: Agnostos web app · Guides: Windows · macOS · Linux · FAQ
An autonomous, token-efficient pipeline for research internship outreach. Agnostos matches your CV against prospective supervisors, retrieves their recent publications, predicts acceptance probability, drafts tailored cold emails without generic LLM prose, and throttles outbound SMTP delivery.
 [code] 
Point it at your CV and a spreadsheet of professors. It reads your resume, ranks the list against it for free, looks up each professor's recent papers, writes a specific email per person, strips the AI tells out of the prose, and sends them on a human-looking schedule — writing every draft and status back into your spreadsheet as it goes.
**It costs nothing.** The default provider is Groq, whose free tier comfortably covers a full run. OpenAI (or any OpenAI-compatible gateway) is supported if you prefer it.
---
## Quick start
 [code] 
Open `.env` and fill in section 1 only: your resume path, your target `.xlsx`, a free Groq key, and your SMTP details. Then:
 [code] 
`stats` reads the workbook (even while it's open in Excel) and every `send` appends each attempt to `output/send_log.csv`, so history survives later edits to the sheet.
Certificate errors (`CERTIFICATE_VERIFY_FAILED ... _ssl.c:...`, common with python.org Python on macOS): run `pip install -U truststore certifi`, then `python -m outreach.cli tls-check`. Behind a TLS-inspecting proxy, set `SSL_CERT_FILE` to its root certificate.
Step-by-step setup guides for each operating system: Windows · macOS · Linux · Questions: FAQ · Something broken: Diagnostics
SMTP is only needed for the last step — you can draft everything without it. Mode A scraping is entirely optional, and `.env.example` is organised so that only its first section is required; everything else has working defaults.
Don't have a list of professors yet? Build one from your research topic instead of collecting names by hand:
 [code] 
See `discover` below.
### Your target spreadsheet
Any `.xlsx` works. The header row is auto-detected and ~25 common column names are mapped, so you usually need nothing beyond:
| Professor Name | Email | Institution | Research Area |
|---|---|---|---|
| Prof Jane Smith | jane.smith@uni.de | TU Munich | Graph neural networks for jet tagging |
`Country`, `Department`, `Fit Reason`, `Funding`, `Eligibility`, and `Application Route` are picked up if present, and any unrecognised column is folded into a notes field the model still sees. Richer research blurbs produce noticeably better drafts.
---
## Commands
### `profile` — verify resume extraction
 [code] 
Prints name, programme, inter