Exa Websets MCP Server

/** * CRM Orchestration: Lead Discovery and Account-Based Marketing * Helps sales and marketing teams find and enrich leads matching their ICP */ export async function crmOrchestration( idealCustomerProfile: string, region?: string, intentSignals?: string ): Promise<string> { const geo = region || "global"; const signals = intentSignals || "hiring, funding, product launches"; return `# 🎯 CRM Orchestration: Lead Discovery and Account-Based Marketing ## What This Solves Discover high-quality leads that match your Ideal Customer Profile (ICP) and enrich them with intent signals, contact information, and engagement history. This orchestration helps sales and marketing teams: - **Lead Discovery**: Find companies matching your ICP criteria - **Intent Scoring**: Identify who's actively buying (hiring, fundraising, launching products) - **Contact Enrichment**: Extract decision-maker names, titles, and email patterns - **Account Deduplication**: Merge duplicate records across sources - **Sales Sequence Planning**: Prioritize outreach based on intent signals ## When to Use ✅ Quarterly lead generation campaigns ✅ Account-based marketing (ABM) targeting ✅ Sales outreach list building ✅ Partner and channel partner discovery ✅ Win/loss analysis and competitive displacement ## 🎯 TL;DR Quick Steps 1. 🔍 **Define & Search**: Search for companies matching your ICP 2. 🧰 **Create Webset**: Organize prospects into a tracked collection 3. ✨ **Enrich**: Extract contact info, intent signals, and company details 4. 📊 **Score**: Rank leads by fit and intent 5. 📧 **Export**: Send to CRM or outreach platform --- ## 📋 Step-by-Step Workflow ### Step 1: Search for ICP-Matching Companies Search for companies that fit your Ideal Customer Profile and show buying intent signals. **Tool**: \`web_search_exa\` **Parameters**: \`\`\`json { "query": "${idealCustomerProfile} (${signals}) location:${geo}", "numResults": 75, "timeRange": "30d", "waitForResults": true } \`\`\` **Expected Output**: 30-75 results with URLs and snippets mentioning companies matching your ICP with intent signals. --- ### Step 2: Create a Lead Webset Create a persistent webset to track and manage qualified leads. **Tool**: \`websets_manager\` **Parameters**: \`\`\`json { "operation": "create_webset_from_search", "name": "ICP Leads - ${idealCustomerProfile} (${geo})", "description": "Leads matching ICP: ${idealCustomerProfile}. Intent signals: ${signals}. Region: ${geo}. Auto-enriched with contact info and intent scoring.", "searchQuery": "${idealCustomerProfile} (${signals}) location:${geo}", "timeRange": "30d", "maxResults": 150 } \`\`\` **Expected Output**: \`\`\`json { "websetId": "ws_leads_icp_456", "status": "processing", "createdAt": "2024-01-10T14:00:00Z", "estimatedCompletion": "2024-01-10T14:30:00Z" } \`\`\` --- ### Step 3: Check Webset Status & Completion Monitor the webset until processing completes. **Tool**: \`websets_manager\` **Parameters**: \`\`\`json { "operation": "get_webset_status", "websetId": "ws_leads_icp_456" } \`\`\` **Expected Output**: \`\`\`json { "websetId": "ws_leads_icp_456", "status": "completed", "itemCount": 142, "completedAt": "2024-01-10T14:28:00Z" } \`\`\` --- ### Step 4: Extract Company & Contact Details Retrieve items and enrich with entity extraction to find company names, decision-makers, and contact patterns. **Tool**: \`websets_manager\` **Parameters** (Get items): \`\`\`json { "operation": "list_items", "websetId": "ws_leads_icp_456", "limit": 100, "offset": 0 } \`\`\` **Expected Output**: Array of 100 lead items with \`url\`, \`title\`, \`text\`, \`publishedDate\`, \`source\`. **Tool**: \`websets_manager\` **Parameters** (Enrich with entity extraction): \`\`\`json { "operation": "enhance_content", "websetId": "ws_leads_icp_456", "enhancements": [ "extract_entities", "summarize", "classify_intent", "extract_contacts" ], "waitForResults": true, "advanced": { "outputFormat": "json", "entityTypes": ["company", "person", "email", "phone"] } } \`\`\` **Expected Output**: Each item enriched with: - \`entities.company\`: ["Acme Corp", "TechCorp Inc", ...]\n- \`entities.person\`: ["John Smith (CEO)", "Jane Doe (VP Sales)", ...]\n- \`entities.email\`: ["john@acme.com", ...]\n- \`summary\`: Key takeaway from the article/page - \`intent\`: "high" | "medium" | "low" - \`intentType\`: "hiring" | "funding" | "product_launch" | "expansion" --- ### Step 5: Score and Tag Leads Add scoring tags to prioritize high-intent, high-fit accounts. **Tool**: \`knowledge_graph\` **Parameters** (Create entities for each company): \`\`\`json { "operation": "create_entities", "entities": [ { "name": "Acme Corp", "entityType": "company", "observations": [ "ICP Fit: High (meets size, industry, tech stack criteria)", "Intent: High (posted 3 job openings this week)", "Decision Makers: John Smith (CEO), Jane Doe (VP Sales)", "Contact Email Pattern: firstname@acme.com", "Website: acme.com" ] } ] } \`\`\` --- ### Step 6: Deduplicate and Export Remove duplicate companies and export to CRM or outreach platform. **Tool**: \`websets_manager\` **Parameters** (Get deduplicated list): \`\`\`json { "operation": "list_items", "websetId": "ws_leads_icp_456", "limit": 150, "offset": 0, "deduplicateBy": "company_domain" } \`\`\` Then export for import into your CRM: **Tool**: \`websets_manager\` **Parameters**: \`\`\`json { "operation": "export_webset", "websetId": "ws_leads_icp_456", "format": "csv", "fields": ["company", "decision_makers", "intent_signal", "contact_email", "website", "source_url"] } \`\`\` **Expected Output**: CSV with columns: - company, decision_makers, intent_signal, contact_email, website, source_url --- ## 💡 Real-World Examples ### Example 1: Mid-Market SaaS Leads with Hiring Intent **Goal**: Find 50-500 person SaaS companies actively hiring (indicating growth/funding). **ICP Definition**: \`"SaaS company" (funding OR "Series A" OR "Series B" OR hiring OR "headcount") employees:50-500\` **Enrichment Focus**: Extract company names, people, hiring announcements **Expected Result**: 30-50 qualified leads with recent hiring posts, founder names, and funding news --- ### Example 2: Enterprise Companies with Product Launch Intent **Goal**: Identify large enterprises launching new products (indicator of project budget). **ICP Definition**: \`"enterprise software" OR "Fortune 500" ("new product" OR "product launch" OR "announcement" OR "press release") industry:technology\` **Enrichment Focus**: Classify intent, extract decision-makers, summarize announcement **Expected Result**: 15-30 hot leads with launch details and likely stakeholders --- ### Example 3: Vertical-Specific Lead Generation (e.g., Healthcare) **Goal**: Find healthcare organizations expanding AI/ML capabilities. **ICP Definition**: \`"healthcare company" OR "hospital system" OR "health network" ("AI" OR "machine learning" OR "analytics" OR "data") hiring:engineering\` **Enrichment Focus**: Extract company size, location, IT leadership, funding status **Expected Result**: 40-80 mid-market healthcare buyers with active AI initiatives --- ## 🎓 Best Practices ### Lead Definition & Targeting - Define ICP with 4-6 core attributes (industry, size, location, technology, growth stage) - Add intent signals to filter: hiring posts, funding announcements, job boards, product launches - Use location/region constraints if relevant - Refine search incrementally (start with 50 results, scale to 150+) ### Enrichment Pipeline - Always run \`extract_entities\` first to get company/people data - Follow with \`classify_intent\` to prioritize high-intent accounts - Use \`summarize\` to create 1-line pitch preparation - Add custom tags in knowledge_graph to categorize (e.g., "hot_lead", "warm", "nurture") ### Deduplication & Quality Control - Export to CSV and run dedup in your CRM to merge duplicate companies - Manually spot-check 10-15 leads for relevance - Remove low-scoring or out-of-scope results before CRM import - Document your ICP and search filters for future campaigns ### Outreach Sequencing - Prioritize high-intent + high-fit leads (top 20%) - Warm up medium-fit leads with content - Save low-fit for future campaigns (nurture list) - Re-run webset monthly to find new leads showing intent ### Multi-Touch Enrichment - Layer multiple websets: one for hiring intent, one for funding, one for job board activity - Use knowledge_graph to connect related signals (same person appearing in multiple contexts) - Track engagement: did this lead respond? When? To what message type? --- ## ⚠️ Troubleshooting ### Issue: Too many/too few results **Cause**: If too few/zero results, likely a technical error (API rate limit, network issue, service unavailability). If too many, query may need refinement. **Solution**: - If zero results: - Check API key is set correctly: \`echo $EXA_API_KEY\` - Verify network connectivity to Exa API - Retry after a few minutes (may be rate limited or service issue) - Contact support with webset ID and timestamp - If too many results: - Add constraints (company size, industry, specific technology) - Narrow intent signals or use more specific phrases ### Issue: Low-quality leads in results **Cause**: Search includes adjacent competitors or non-decision-makers. **Solution**: - Add negative filters: -"competitor name", -"media coverage", -"job board" - Increase intent signal specificity - Manually review and refine ICP definition ### Issue: Can't find contact information **Cause**: Articles/sources don't mention email addresses directly. **Solution**: - Use \`extract_entities\` with \`includeEmails: true\` - Follow up with LinkedIn search for decision-makers found by name - Check company website domain and create email pattern (firstname@domain.com) ### Issue: Enrichment job times out **Cause**: Webset too large (500+ items). **Solution**: - Split into smaller websets (200 items max) - Prioritize top-scoring items first - Run enrichment in batches ### Issue: Duplicate companies in export **Cause**: Same company mentioned in multiple sources. **Solution**: - Use \`deduplicateBy: "company_domain"\` when listing items - Run dedup in your CRM on company domain/name field before import --- ## ❓ FAQ **Q: How often should I run lead generation campaigns?** A: Monthly for active markets. Quarterly for stable verticals. Re-run when ICP changes or new signals emerge. **Q: Can I combine multiple search queries in one webset?** A: Yes! Use OR logic in your search: \`(ICP_criteria_1) OR (ICP_criteria_2)\` **Q: How do I avoid spam/low-quality leads?** A: Filter by credibility indicators—news sites, company blogs, press releases vs. forums. Increase intent signal specificity. **Q: Can I track which leads become customers?** A: Yes! Tag leads in knowledge_graph with your CRM ID, then track conversions back to source webset. **Q: How long does this take?** A: Webset creation: 10-30 min. Enrichment: 5-15 min. Export: <1 min. End-to-end: 30-60 min for 100-150 leads. **Q: What if I want to nurture leads instead of hard-sell?** A: Create a separate "nurture" webset with lower-intent signals (mentions of your industry, early-stage keywords), and export as a content audience list. --- ## 🚀 Next Steps 1. Define your ICP with concrete attributes and intent signals 2. Create your first lead webset (aim for 50-150 leads) 3. Enrich with entity extraction and intent classification 4. Deduplicate and manually review top 20% of leads 5. Export to CRM and set up outreach sequence 6. Track which webset-sourced leads convert Ready to discover your next 100 qualified leads?`; }