eracle/OpenOutreach

OpenOutreach – Self‑hosted AI lead‑generation agent

What it is – An open‑source command‑line tool that lets you describe your product and target market, then uses a language model and a licensed lead‑data service to discover, qualify, and email prospects. All processing runs on your own machine; the only external calls are to an LLM API (e.g., OpenAI or Anthropic) and to the BetterContact email‑finder service.

Who it’s for – Small‑to‑medium B2B teams, founders, or developers who want automated outbound outreach without paying a SaaS cold‑email platform or uploading a pre‑made list.

Key features

• One‑command onboarding – uv tool install openoutreach && openoutreach walks you through four inputs (LLM key, BetterContact key, product description, target market) and stores them in ~/.openoutreach.
• Lead discovery & qualification – The companion OpenOutFind searches a licensed data source, the LLM turns your description into search keywords, and a small active‑learning loop decides which candidates are a fit, attaching a human‑readable reason for each lead.
• Email enrichment & sending – OpenOutSend resolves a verified work email (paid per lookup) and sends a personalized opener from your own mailbox, respecting pacing and daily caps.
• CLI‑only, no browsers or social‑media accounts – No scraping, no platform‑ToS risk.
• Export‑ready CSV – openoutreach find 10 emails > leads.csv produces a file with columns that match common sequencer imports (email,first_name,last_name,company,title,website,linkedin_url,reason,…).
• Claude Code integration – A ready‑made Claude plugin/skill lets any LLM‑driven agent invoke the same commands.

How it works (high‑level)

1. You give a short product + market description.
2. The LLM generates search keywords.
3. OpenOutFind queries BetterContact’s licensed index (free) for matching firms.
4. An LLM evaluates each candidate against your Ideal Customer Profile, writes a plain‑language reason, and a Gaussian‑process model learns from those verdicts to prioritize the next batch.
5. When the confidence gate is passed, a paid BetterContact lookup fetches a verified work email (one credit per hit).
6. OpenOutSend composes an opener using the same product description and sends it via your SMTP credentials.

Installation & quick start

uv tool install openoutreach   # installs the package and its two child CLIs
openoutreach                  # runs the wizard, then finds & emails leads
# or more explicit
openoutreach run 5            # find up to 5 qualified leads with emails and send them

All state lives under ~/.openoutreach; re‑running the command continues where it left off.

Core commands

• openoutreach – full wizard + find + send (default).
• openoutreach init – only run the onboarding wizard.
• openoutreach find <n> [emails] – fetch n qualified leads; add emails to spend credits for verified addresses.
• openoutreach send [<n>] – send emails for stored leads, optionally limiting to n.
• openoutreach status – show configuration, credit usage, and queued leads.

Dependencies

• Python (installed via uv or pip).
• An LLM API key (OpenAI, Anthropic, or compatible).
• A BetterContact API key (free tier provides 40 email‑lookup credits).
• SMTP credentials for the mailbox you’ll send from.

Project layout

• openoutreach/ – orchestrator code (__main__.py, settings, wizard).
• Child projects (OpenOutFind, OpenOutSend) are pulled in as pinned dependencies and can be used separately.
• Tests, Docker compose (local.yml), and Makefile for development.

License & sustainability

• Licensed under GPL‑v3.
• The tool stays free by earning affiliate commissions from the BetterContact sign‑up link; you may use any other provider if you prefer.

Where to get help

• Telegram channel: https://t.me/openoutreach
• Issues on GitHub, or book a free 15‑minute call via the badge in the README.

---

All details are taken directly from the repository’s README; no additional features are inferred.