Metadata-Version: 2.4
Name: superpitch
Version: 0.3.1
Summary: AI screening agent for investor inbound. Processes pitch emails locally, learns from every decision.
License: MIT
Requires-Python: >=3.10
Description-Content-Type: text/markdown
License-File: LICENSE
Provides-Extra: specialist
Requires-Dist: mlx-lm; extra == "specialist"
Dynamic: license-file

# Superpitch

[![PyPI](https://img.shields.io/pypi/v/superpitch)](https://pypi.org/project/superpitch/)
[![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](LICENSE)

AI screening agent for your inbound pitch emails.

---

## Install

```bash
curl -LsSf https://superpitch.dev/install.sh | sh
```

Then run `superpitch`.

---

## Setup

1. **Sign in** — run `superpitch`, type `/signin`. You get a private address like `a7f3k2@superpitch.dev`.
2. **Forward pitches** — set a Gmail or Outlook rule to copy inbound pitch emails to that address.
3. **Set your thesis** — `/thesis`. A short description of what you invest in.

That's it. Open `superpitch` and your inbox is screened.

---

## How it works

A pitch email arrives at your private address. Within minutes:

1. **Noise filter** — the agent classifies the email instantly. Newsletters, sales emails, and non-pitches are discarded. Only real pitches continue.

2. **Deep research** — the agent reads the full email, any attached PDF deck, and searches the web for the company and founder. It pulls public data — LinkedIn, Crunchbase, press, product pages.

3. **Evidence packet** — everything is structured into a single packet: company overview, founder background, stage, sector, traction, raise details, business model, thesis fit assessment, research findings with sources, and sharp questions worth asking before a call.

4. **Triage** — each pitch is classified as ESCALATE, REVIEW, or PASS with a confidence score based on your thesis.

5. **Founder follow-up** — when a pitch is promising but incomplete, the agent emails the founder directly with targeted questions. The founder replies, the agent reads the answer, and the evidence packet updates — as many rounds as needed until the packet is complete. All before you've opened the app.

You open `superpitch` and everything is waiting, ranked and ready.

---

## What you see

```
  Superpitch  Logged in as you@email.com              2 min ago
  ──────────────────────────────────────────────────────────────

  Inbox
  ······························································
    DATE        COMPANY                  ACTION     CONF   STATUS
  ······························································

    2026-03-20  Mira Health              ESCALATE   91%       New
    2026-03-20  Stackwise                REVIEW     74%  Awaiting
    2026-03-19  LogiRoute                REVIEW     68%    Opened
    2026-03-18  Widget Co                PASS       31%       New

  ──────────────────────────────────────────────────────────────
  ↑↓ move  ·  ↵ open  ·  a/r/v decide  ·  / commands  ·  q quit
```

---

## Your data stays local

Your decisions, history, and screening preferences live on your machine. The agent learns from every approve, reject, and review — a local model retrains in the background after every 20 decisions, adapting to your judgment over time. No data leaves your device for training.

---

## One-off analysis

```bash
superpitch pitch.pdf
```

Drop a file and get the evidence packet printed directly. No email forwarding needed.

---

## Pricing

**$299/month** or **$249/month billed yearly**.

Everything included — AI screening, web research, PDF deck reading, automatic founder follow-up, personal pitch URL, and the terminal inbox.
