FameClaw

Open-source YouTube creator outreach — from finding creators to closing deals.

100% local. Your data, emails, and credentials never leave your machine. Nothing is sent to any server, ever. No API keys, no cloud services, no tracking. Just curl + python3.

What It Does

1. Scan your brand's website → auto-detect product category & niche
2. Search YouTube for creators in matching niches
3. Extract channel stats: subscribers, views, emails, descriptions
4. Discover related channels via YouTube's recommendation algorithm
5. Score every channel against your brand's audience profile (0-100)
6. Email the best matches — personalized, mentions their actual videos
7. Follow up automatically — day 3 and day 8 if no reply
8. Negotiate autonomously — reads replies, counters, closes deals

Privacy

• All data stays on your device — CSVs, emails, credentials, campaign state
• Gmail credentials stored locally (~/.config/fameclaw/gmail.json, mode 600)
• No external APIs, no cloud services, no telemetry, no tracking
• Emails sent directly from your mailbox via SMTP — no middleman
• Works with Gmail, Outlook, iCloud, Yahoo, Zoho, Fastmail, or any SMTP server
• Open source — read every line of code yourself

Quick Start

Single channel

bash scripts/extract_channel_data.sh "https://youtube.com/@handle" output.csv

Batch prospecting

bash scripts/prospect.sh \
  --queries "tiktok shop tutorial" "dropshipping beginner" \
  --target 100 \
  --output channels.csv \
  --max-subs 100000

Full pipeline with audience matching

# 1. Scan brand site
bash scripts/onboard.sh --brand "MyBrand" --url "https://mybrand.com" --output scan.json

# 2. Create audience profile (edit the generated template)
# Sets target categories, demographics, authority preferences

# 3. Prospect
bash scripts/prospect.sh --config config.json

# 4. Score & rank
python3 scripts/score_channels.py --csv channels.csv --profile audience.json --output scored.csv

Audience Matching

FameClaw scores creators using three dimensions:

Dimension	Points	What It Checks
Category match	0-50	Does the channel's content match your niche?
Demographic match	0-30	Does the creator match your target customer?
Authority match	0-20	Does the creator have professional credibility?

14 content categories: Beauty & Skincare, Fitness & Health, Tech & Gadgets, Fashion & Lifestyle, Food & Cooking, E-commerce & Business, TikTok & Social Media, Finance & Investing, Gaming, Education & Tutorial, Home & DIY, Parenting & Family, Pets & Animals, Travel & Outdoor

Match types:

• demographic — creator looks like your buyer (age, gender, lifestyle match)
• authority — creator has credibility (doctor, coach, certified expert, founder)
• demographic+authority — both signals present

Gmail Outreach

Send personalized emails to scraped creators using Google's official Workspace CLI.

Setup (one time)

# 1. Google Account → Security → 2-Step Verification → ON
# 2. Google Account → Security → App passwords → Generate
# 3. Create gmail_creds.json:
echo '{"email": "you@gmail.com", "app_password": "xxxx xxxx xxxx xxxx", "display_name": "Your Name"}' > gmail_creds.json

# 4. Test connection
python3 scripts/gmail.py test --creds gmail_creds.json

Configure campaign

{
  "brand": "MyBrand",
  "website": "https://mybrand.com",
  "sender_name": "Alex",
  "gmail_creds": "gmail_creds.json",
  "current_partnerships": ["@CreatorA", "@CreatorB", "@CreatorC"],
  "rate": 30,
  "min_score": 25,
  "max_per_run": 50
}

Run outreach

# Send first emails (auto-fetches recent videos per creator)
python3 scripts/outreach.py send --csv scored.csv --config outreach.json --dry-run
python3 scripts/outreach.py send --csv scored.csv --config outreach.json

# Check for replies → moves responders to NEGOTIATE
python3 scripts/outreach.py check-replies --config outreach.json

# Auto follow-ups (day 3 + day 8)
python3 scripts/outreach.py followup --config outreach.json

# Campaign dashboard
python3 scripts/outreach.py status --config outreach.json

Email sequence

Three auto-generated emails per creator:

1. Initial — short, mentions their specific video by name, drops 2-3 current partnerships
2. Follow-up (day 3) — bump, references a different video
3. Final (day 8) — last touch, respects their time

All emails stop automatically when the creator replies.

Scripts

Script	Purpose
onboard.sh	Scan a brand's website, extract niche signals
scan_site.py	Site intelligence extraction (title, description, industry, social links)
prospect.sh	Batch prospecting pipeline with cron support
extract_channel_data.sh	Single channel → stats + email → CSV row
extract_email.sh	Email-only extraction from a channel
find_related_channels.sh	Discover related channels via YouTube recommendations
score_channels.py	Score & rank channels against audience profile
outreach.py	Multi-stage outreach pipeline (send, follow-up, reply detection)
gmail.py	Gmail client — SMTP send + IMAP reply tracking
outreach.sh	Simple one-shot email sender (legacy, uses gws CLI)
get_videos.sh	Fetch recent video titles for personalization
negotiate.py	Autonomous negotiation engine (classify, counter, close)

Config

{
  "queries": ["tiktok shop affiliate tutorial", "faceless youtube channel"],
  "target_emails": 100,
  "output": "channels.csv",
  "max_subs": 100000,
  "batch_size": 200,
  "work_dir": "./prospect-run",
  "cron_name": "my-prospect-job"
}

Audience Profile

{
  "brand": "MyBrand",
  "url": "https://mybrand.com",
  "target_categories": ["Beauty & Skincare", "Fashion & Lifestyle"],
  "target_demographics": {
    "age_range": "25-40",
    "gender": "female",
    "interests": ["skincare", "clean beauty"],
    "location": "US"
  },
  "authority_preferred": true
}

CSV Output

channel_name, handle, subscribers, total_videos, avg_views, median_views,
min_views, max_views, videos_sampled, email, description, external_links,
channel_url, content_category, match_score, match_type, match_reasons

Cron Automation

For large targets, run on a schedule:

# Using OpenClaw
openclaw cron add --name "my-prospect" --every 30m \
  --session isolated --model haiku --timeout-seconds 1800 \
  --message "Run: bash scripts/prospect.sh --config config.json"

# Using system cron
*/30 * * * * bash /path/to/scripts/prospect.sh --config /path/to/config.json

The script auto-removes the cron job when the email target is reached.

How Email Discovery Works

1. Scan YouTube channel page for emails in metadata/description
2. Derive vanity domain from handle (strips "live", "official", "hq")
3. Check vanity domain: root, /contact, /about, /contact-us
4. Filter junk (image files, google/youtube domains, noreply)

Requirements

• curl
• python3 (3.9+)
• No API keys needed
• OpenClaw or NanoClaw (optional, for AI agent integration)

AI Agent Integration

FameClaw works as a skill for both major personal AI agent platforms:

OpenClaw

Drop the skill into your OpenClaw workspace and the agent handles the entire pipeline conversationally — onboarding, prospecting, scoring, outreach. Uses OpenClaw's cron system for automated runs.

NanoClaw

FameClaw ships with a NanoClaw-compatible skill in the nanoclaw/ directory. Copy it to your NanoClaw's .claude/skills/fameclaw/ directory:

cp -r nanoclaw/ /path/to/nanoclaw/.claude/skills/fameclaw/

The scripts run inside NanoClaw's container sandbox. Mount ~/.config/fameclaw/ for credentials and your working directory for CSV output. Uses NanoClaw's task scheduler for recurring runs.

Both integrations use the same core scripts — zero divergence, zero API keys, fully local.

License

MIT — most of FameClaw is fully open source.

FSL-1.1-MIT — the scoring engine (score_channels.py) and negotiation engine (negotiate.py) are licensed under the Functional Source License. You can use and modify them for any purpose except building a competing product. Redistribution is not permitted. Companies under $1M/year revenue can use them freely, including commercially. Companies over $1M/year need a commercial license.