Outreach & Lead Generation Guide

The primary entry point for any outreach task. find_leads takes a natural-language goal and decomposes it into directory searches, corporate-site parses, and social-network probes. It runs up to 6 corporate-site parses per company in parallel and merges results on the fly.

Parameters

Example invocation

result = await agent.call_tool("find_leads", {
  industry: "B2B SaaS",
  location: "San Francisco Bay Area",
  roles: ["VP Engineering", "CTO"],
  limit: 25,
  enrich: true
})

Example output (one record)

{
"company": "Acme Analytics",
"website": "https://acme-analytics.com",
"industry": "B2B SaaS",
"location": "San Francisco, CA",
"employees": "80-120",
"funding_stage": "Series B",
"contacts": [
  {
    "name": "Jordan Lee",
    "title": "VP Engineering",
    "email": "[email protected]",
    "email_confidence": 0.97,
    "linkedin": "https://linkedin.com/in/jordanlee",
    "source_url": "https://acme-analytics.com/team"
  }
],
"confidence": 0.92,
"sources": [
  "https://2gis.com/san-francisco/firm/acme-analytics",
  "https://acme-analytics.com/team",
  "https://linkedin.com/company/acme-analytics"
]
}

Searches across 2GIS, Google Maps, Yandex Maps, and Yellow Pages in a single call. The tool fans out to all available directories for the target geography, deduplicates by normalized name

• address, and returns a unified result set.

Parameters

Returned fields per record

• name — Normalized business name
• category — Business category from the directory
• address — Full street address
• phone — Phone number in E.164 when available
• website — Company website URL
• email — Public email if listed in the directory
• rating — Average star rating (Google Maps, 2GIS)
• source — Which directory the record came from

Example output

{
"name": "NovaTech Solutions",
"category": "Software Development",
"address": "45 Market St, Suite 200, San Francisco, CA 94105",
"phone": "+14155550123",
"website": "https://novatech.io",
"email": "[email protected]",
"rating": 4.6,
"reviews_count": 38,
"source": "google_maps"
}

Searches Twitter/X, Telegram, and LinkedIn for people and companies matching a query. Useful for finding decision-makers by title+company, discovering niche community participants, and cross-referencing names found through directory searches.

Parameters

Returned fields per record

• profile_url — Direct link to the profile
• name — Display name
• bio — Profile bio / description text
• followers_count — Follower count (Twitter, LinkedIn approximate)
• location — Self-reported location
• platform — Source platform identifier

Example output

{
"profile_url": "https://twitter.com/jordanlee_eng",
"name": "Jordan Lee",
"bio": "VP Engineering @AcmeAnalytics · Building data pipelines at scale",
"followers_count": 4200,
"location": "San Francisco, CA",
"platform": "twitter"
}

Searches news articles, press releases, and blog coverage. Especially useful for identifying recently funded companies, new executive hires, product launches, and partnership announcements — all high-signal events for outreach timing.

Parameters

Returned fields per record

• title — Article headline
• url — Canonical URL
• source — Publication name
• date — Publication date
• snippet — 200-character excerpt
• mentioned_persons — Extracted named entities (people)
• mentioned_companies — Extracted named entities (companies)

Example output

{
"title": "Acme Analytics Raises $40M Series B Led by Sequoia",
"url": "https://techcrunch.com/2025/03/15/acme-series-b",
"source": "TechCrunch",
"date": "2025-03-15",
"snippet": "Acme Analytics, the San Francisco-based B2B data platform, has closed a $40M Series B round...",
"mentioned_persons": ["Jordan Lee", "Maria Chen"],
"mentioned_companies": ["Acme Analytics", "Sequoia Capital"]
}

Analyzes a corporate website to extract structured company intelligence. Crawls the homepage, /about, /team, and /contact pages (following robots.txt). Resolves obfuscated emails and phone numbers embedded in JavaScript or rendered via anti-scraping techniques.

What gets extracted

• Company name — From <title>, OG tags, or structured data (JSON-LD)
• Description — Meta description or first meaningful paragraph
• Industry — Inferred from content keywords + JSON-LD categories
• Size — Employee count range from careers page or LinkedIn widget
• HQ — Address from footer, contact page, or structured data
• Contact emails — Extracted from pages, mailto links, and JS-rendered content
• Contact phones — Parsed from all pages, including international formats
• Social profiles — Twitter, LinkedIn, GitHub, Telegram links from headers/footers
• Team members — Names and titles from /team, /about, or leadership pages

Example output

{
"company_name": "Acme Analytics",
"description": "Real-time data pipelines for modern B2B teams",
"industry": "B2B SaaS / Data Infrastructure",
"size": "80-120 employees",
"hq": "45 Market St, Suite 200, San Francisco, CA 94105",
"emails": ["[email protected]", "[email protected]"],
"phones": ["+14155550100"],
"social": {
  "twitter": "https://twitter.com/acmeanalytics",
  "linkedin": "https://linkedin.com/company/acme-analytics",
  "github": "https://github.com/acme-analytics"
},
"team": [
  { "name": "Jordan Lee", "title": "VP Engineering" },
  { "name": "Maria Chen", "title": "CEO" },
  { "name": "Alex Rivera", "title": "CTO" }
],
"technologies": ["React", "Go", "Kubernetes", "PostgreSQL"]
}

The most critical verification tool. Every email passes through five sequential stages — each stage can short-circuit the pipeline if a hard failure is detected (e.g., invalid syntax). The final confidence score is a composite of all stage results.

The 5-stage pipeline in detail

Confidence scoring algorithm

score = 0.5                                    # base
if mx_resolved:     score += 0.4               # domain is real
if disposable:      score -= 0.3               # throwaway domain
if role_based:      score -= 0.15              # department mailbox
if smtp_accepted:   score = max(score, 0.95)   # server confirmed
if smtp_rejected:   score = min(score, 0.2)    # server denied
if no_smtp_check:   score = min(score, 0.9)    # can't fully confirm
score = clamp(score, 0.0, 1.0)
if pattern_matched: score = min(score + 0.15, 0.95)  # suggest_emails bonus

Example responses

{
"email": "[email protected]",
"verdict": "valid",
"confidence": 0.97,
"stages": {
  "syntax": "pass",
  "mx": "pass",
  "disposable": "clean",
  "role_based": false,
  "smtp": "accepted (250 OK)"
},
"mx_records": ["alt1.gmail-smtp-in.l.google.com", "gmail-smtp-in.l.google.com"]
}

{
"email": "[email protected]",
"verdict": "disposable",
"confidence": 0.05,
"stages": {
  "syntax": "pass",
  "mx": "pass",
  "disposable": "flagged (mailinator)",
  "role_based": false,
  "smtp": "skipped"
}
}

{
"email": "[email protected]",
"verdict": "role_based",
"confidence": 0.78,
"stages": {
  "syntax": "pass",
  "mx": "pass",
  "disposable": "clean",
  "role_based": "flagged (info@)",
  "smtp": "accepted (250 OK)"
}
}

Normalizes phone numbers using libphonenumber (the same library behind Google’s Android dialer). Parses raw input into E.164 format, detects the country, and classifies the line type.

What it returns

• e164 — E.164 canonical format (e.g. +14155550123)
• country — ISO-3166-1 alpha-2 country code
• line_type — mobile, fixed_line, voip, toll_free, or unknown
• valid — Boolean: does the number match the numbering plan for the detected country?
• carrier — Carrier name when available (mobile numbers only)

Example

{
"input": "8 (495) 123-45-67",
"e164": "+74951234567",
"country": "RU",
"line_type": "fixed_line",
"valid": true,
"carrier": null
}

Validates and enriches social profile URLs. Returns structured metadata specific to each platform — follower counts, bios, verification status, and more.

Supported platforms (14)

Example response

{
"url": "https://twitter.com/jordanlee_eng",
"platform": "twitter",
"valid": true,
"metadata": {
  "name": "Jordan Lee",
  "handle": "@jordanlee_eng",
  "bio": "VP Engineering @AcmeAnalytics",
  "followers": 4200,
  "following": 890,
  "verified": false,
  "joined": "2019-04-22",
  "location": "San Francisco, CA"
}
}

Given a person’s full name and a company domain, generates candidate email addresses using the 9 most common corporate patterns. When known addresses for the same domain are available, it infers the company’s actual pattern and boosts matching candidates.

The 9 standard patterns

Confidence scoring for suggestions

• Each candidate gets a base confidence of 0.3 — it’s just a pattern guess.
• If the inferred pattern from known addresses matches → +0.15 (capped at 0.95).
• Running verify_email on the suggestion can push the score up to 0.97.
• Suggestions that fail MX lookup drop to 0.05 immediately.

Takes a company name or domain and returns structured intelligence gathered from public sources: corporate filings, about pages, job postings, technology stacks, and press coverage.

What gets discovered

• Industry — Primary and secondary SIC/NAICS categories
• Company size — Employee range (e.g. 80-120)
• Revenue — Estimated annual revenue when publicly available
• Founding year — Year of incorporation or launch
• HQ — Headquarters address
• Description — One-paragraph company description
• Technologies — Detected tech stack from job postings, headers, and scripts
• Funding — Latest funding round, amount, and investors

Example response

{
"domain": "acme-analytics.com",
"company_name": "Acme Analytics",
"industry": "B2B SaaS — Data Infrastructure",
"employee_range": "80-120",
"estimated_revenue": "$10M-$25M",
"founded": 2019,
"hq": "San Francisco, CA, USA",
"description": "Real-time data pipeline platform for B2B analytics teams.",
"technologies": ["React", "Go", "Kubernetes", "PostgreSQL", "Snowflake"],
"funding": {
  "latest_round": "Series B",
  "amount": "$40M",
  "date": "2025-03-15",
  "investors": ["Sequoia Capital", "Y Combinator"]
}
}

Takes a name and company (or email) and discovers publicly available professional information. Cross-references LinkedIn, Twitter, GitHub, press mentions, and conference speaker lists.

What gets discovered

• Title — Current job title
• Company — Current employer
• LinkedIn — Profile URL
• Twitter / X — Profile URL and handle
• GitHub — Profile URL and notable repos
• Public contact details — Emails and phones found in public directories
• Location — City / metro area
• Bio — Summary from LinkedIn or conference bios

Example response

{
"name": "Jordan Lee",
"title": "VP Engineering",
"company": "Acme Analytics",
"location": "San Francisco, CA",
"bio": "Engineering leader with 12 years in data infrastructure. Previously at Stripe and Google.",
"linkedin": "https://linkedin.com/in/jordanlee",
"twitter": "https://twitter.com/jordanlee_eng",
"github": "https://github.com/jordanlee",
"emails": ["[email protected]"],
"phones": ["+14155550199"],
"recent_mentions": [
  {
    "title": "Acme Analytics Raises $40M Series B",
    "source": "TechCrunch",
    "date": "2025-03-15"
  }
]
}