sellbird README

The project's README from github.com/ethanplusai/sellbird. As one markdown file: README.md.

Sellbird

The open-source outreach platform that runs on your Claude or ChatGPT plan.

Sellbird finds companies that need what you sell, finds the person who decides, checks their email address, writes each one a personal cold email, sends it and handles the replies. It runs on your own machine, and by default nothing goes out until you approve it.

GitHub stars Tests License: AGPL-3.0 Python 3.11+ Hosted at sellbird.ai Formerly Harvey

Harvey is now Sellbird. Same project, same history, new name. Already running Harvey? Your setup carries over: see Upgrading from Harvey.

Prefer not to run it yourself? sellbird.ai is the hosted version, with sending domains, inboxes and warmup set up for you. It's in early access.

Sellbird's dashboard: the Today screen, a conversation that needs you, approving a drafted reply in the Outbox, the signals to confirm, searching contacts, then dark mode

The demo data is invented: "Bookwell" selling online booking to roofers, dental practices and HVAC shops, on .test domains.

What it does

Sellbird is a pipeline you can run end to end, or stop at any step.

What happens More
Find companies Searches for businesses that match your customer: free on OpenStreetMap, or on a paid source you pick. Every paid search shows its cost before it runs. Discovery
Learn why they'd buy Reads each company's own website for the buying signals you confirmed, like no online booking, ads running or a role they're hiring for. No model calls, no cost. Signals
Find who decides Finds the owner or manager named on the company's own pages, two people per company at most. Enrich
Check the address Learns the company's email pattern and verifies one address. Every address is marked verified, risky, guess or invalid, and only deliverable ones are ever sent to. Verify
Write and send Writes a short sequence for each person from what it found, then sends from your Gmail, Microsoft 365 or SMTP mailbox inside your daily limit and their office hours. Sending
Handle replies Reads every reply, drafts an answer, sets auto-replies aside, honours opt-outs at once, and hands you anything it shouldn't answer alone. Dashboard

Only want the list? Stop after verifying and run sellbird export for a sequencer-ready CSV of everything it found.

$ sellbird run

============================================================
Sellbird is online.
============================================================
Checking pipeline state...
Decision: write_campaign (3 new prospects with no drafts)
  handler   → Elena Park, Harbor Dental Studio     interested, answer drafted
  handler   → Priya Nair, Highline Roofing         objection, answer drafted
  writer    → 2-email sequence: Phoenix HVAC hiring a dispatcher
  outbox    → 5 emails waiting for your approval
Cycle complete. Sleeping for 15 minutes.

No model API bill

Sellbird thinks through the Claude Code CLI on your Claude Pro or Max plan, or through Codex on a ChatGPT plan. There's no API key and no per-token billing, and Sellbird throttles itself to leave room for your own work.
Claude Pro or Max, or a ChatGPT plan with Codex the subscription you already have
Finding businesses $0 on the default source, or about $0.37 per 1,000 on the cheapest paid one
Reading their websites $0: three HTTP requests each, no model call
Writing and sending your subscription and a mailbox (about $7 a month for Google Workspace)
Prospecting tools None. No Apollo, ZoomInfo, Clearbit or Clay.

How the loop, the budget and the isolation work: How Sellbird works.

Quick start

You need:

Let Claude set it up

git clone https://github.com/ethanplusai/sellbird.git
cd sellbird
claude

Then say "set up Sellbird for me". The repo's CLAUDE.md turns Claude Code into the setup guide: it checks where you are, installs what's missing, asks what it needs and trains Sellbird on your product.

Or do it yourself

# 1. Install. Use python3.13, not python3: on macOS bare python3 is 3.9.
python3.13 -m venv .venv && source .venv/bin/activate && pip install -e .

# 2. Add credentials. Every variable is documented inline.
cp .env.example .env && $EDITOR .env

# 3. Learn your product from your website (writes sellbird.local.yaml).
sellbird train https://your-company.com

# 4. Confirm what makes a good prospect.
sellbird signals --confirm free

# 5. Find businesses. The default source is free and needs no account.
sellbird discover

# 6. Open the dashboard and start the loop.
sellbird dashboard          # prints a private link to http://127.0.0.1:5555
sellbird run

On Linux you can also run it with Docker.

Nothing sends without you

Every outgoing email waits in the Outbox for your approval until you turn on autopilot yourself. Before anything leaves, a pre-send check with no model involved rejects unfilled merge tags, banned phrases, overlong emails, extra links and undeliverable addresses. Autopilot still holds back any email that links to a domain that isn't yours or quotes Sellbird's own instructions. sellbird sending pause stops everything at once.

Details: Approval and sending and Legal and deliverability.

Upgrading from Harvey

Harvey was renamed Sellbird in 0.4.0. Your configuration, data and dashboard sign-in carry over:

git pull
pip install -e .        # once, to get the new `sellbird` command
sellbird dashboard

Links to github.com/ethanplusai/harvey keep working: GitHub redirects them here.

Where your files live

Your config, knowledge and data live in one home folder. Which one, first match wins:

  1. SELLBIRD_HOME, if you set it (the Docker image sets it to /app).
  2. A checkout: the checkout folder itself, as before.
  3. Otherwise, for an installed package (pip or pipx), your user data folder:
Home folder
macOS ~/Library/Application Support/sellbird
Windows %APPDATA%\sellbird
Linux ~/.local/share/sellbird (or $XDG_DATA_HOME/sellbird)

Inside it are sellbird.yaml (and sellbird.local.yaml once you train), .env, prompts/ and skills/, all yours to edit. An installed package copies the defaults in on first run and never overwrites them, so your edits survive upgrades. data/ holds the database, the logs and the dashboard token.

Documentation

Guide What's in it
How Sellbird works The heartbeat loop, the subscription budget, project structure and the ideas behind it
Signals and the prospecting pipeline The 29 signals, cohorts, and the discover, profile, enrich and verify stages
Skills and sub-agents The editable sales playbooks and which agent uses which
Approval and sending The Outbox, autopilot, office hours and email signatures
The dashboard and the data Every screen, how sign-in works, and how the data is stored
Configuration and commands .env, sellbird.yaml and every command
Extensions Adding routes, heartbeat steps and commands from a separate package
Using a ChatGPT plan (Codex) Running on Codex instead of Claude
Legal and deliverability The rules you're sending under, and warmup
Microsoft 365 · Precleared Provider setup guides
Docker · Troubleshooting · Roadmap Running, fixing and what's next

Contributing

Pull requests are welcome. Read CONTRIBUTING.md first: it covers keeping real contact data out of the repo, test email, and signing off your commits. Found a security problem? Follow SECURITY.md instead of opening an issue.

See CHANGELOG.md for what changed in each release.

License

Sellbird is licensed under the GNU Affero General Public License v3.0. Harvey releases up to 0.3.x were released under MIT, and that code keeps its MIT notice in NOTICE.

TrulyInbox links in this project are affiliate links: the maintainer may earn a commission, at no extra cost to you.

sellbird (ethanplusai/sellbird)

Searching Database…

4.6K ·GitHub
Sellbird
What is this?Harvey is an autonomous AI sales agent that finds businesses to sell to, learns something specific about each one, writes a personalized cold email, sends it, reads replies, and pushes the conversation toward a booked meeting — running in a loop on your own machine.
Is this a hosted SaaS or something I run myself?It is a self-hosted Python program: you run it on your own machine with your own login, and it exposes a local web dashboard at localhost:5555 rather than a cloud account.
How do I get it running?Install Python 3.11+, log in to the Claude Code CLI (or set up Codex), and tell Claude Code to 'set up Harvey for me' — the repo ships a CLAUDE.md that acts as a setup wizard, checking your state, installing what's missing, and training Harvey on your product.
Do I need a Claude or OpenAI API key to run it?No separate API key — Harvey uses your existing Claude Pro/Max subscription through the Claude Code CLI (default) or a ChatGPT plan through Codex (opt-in on macOS/Linux), plus a mailbox like Gmail/Google Workspace to actually send mail from.
Does Harvey send cold emails on its own without me checking them?No — the project explicitly guarantees that nothing sends without you; drafted outreach is approved message-by-message from the dashboard or terminal before it goes out.
Does Harvey only work on macOS, or Linux too?The Claude provider is the default on any machine with Python 3.11+ and the Claude Code CLI; the Codex (ChatGPT) provider is explicitly supported as an opt-in option on macOS and Linux.
How much does finding prospects actually cost?The default source, OpenStreetMap, is free with unlimited lookups but thin coverage for storefront-less trades; paid sources like DataForSEO or Serper cost roughly $0.0018–$1 per search or per 1,000 listings.
What signals does Harvey check before deciding who to target?It can track 23 signals (no website, unclaimed Google listing, running ads, stale blog, missing schema markup, etc.) but collects none of them until you review and confirm the catalog yourself.
How do I change the way Harvey writes its emails?Harvey's sales knowledge lives as plain Markdown files in skills/ (like emailframeworks.md or objectionhandling.md); editing a file changes behavior on the next 15-minute heartbeat with no retraining or restart needed.
Why doesn't Harvey use tools like Apollo or Clearbit?It replaces paid prospecting databases entirely, building contact and company data itself from free sources and three plain HTTP requests per business (homepage, robots.txt, sitemap) instead of a subscription data provider.
What are similar projects?franco, brigade, OpenDuet, autoresearch, hermes-agent, buscandomilhao, growth-engineer, CUAWright
Which topics is it in?sales-automation, lead-generation, email-marketing, autonomous-agent, email, autonomous-agents, chatgpt, workflow-automation, agentic-ai, dashboard, claude, openai
Creator
ethanplusai
Size
17.1 MB
Stars
4.6K
Updated
CommitNov 06Releasev0.4.0Nov 06
Age
7mo
Commits · 90 Days