Exa Websets MCP Server

/** * Marketing Orchestration: Competitor and Brand Monitoring * Helps teams discover brand mentions, track competitors, and monitor industry trends */ export async function marketingOrchestration( companyName: string, targetAudience?: string, timeframe?: string ): Promise<string> { const audience = targetAudience || "general market"; const period = timeframe || "30d"; return `# 📈 Marketing Orchestration: Competitor and Brand Monitoring ## What This Solves Track your brand's presence online, monitor competitor activity, and discover emerging industry trends. This orchestration helps marketing teams: - **Share of Voice**: Monitor mention frequency vs. competitors - **Brand Health**: Track sentiment and perception shifts - **Content Gaps**: Identify topics your competitors cover but you don't - **Influencer Discovery**: Find who's talking about your space - **Market Intelligence**: Stay ahead of industry movements ## When to Use ✅ Weekly/monthly brand health checks ✅ Competitive win/loss analysis ✅ Crisis management and reputation monitoring ✅ Content strategy planning ✅ Market positioning research ## 🎯 TL;DR Quick Steps 1. 🔍 **Search**: Find brand mentions and competitor references across the web 2. 🧰 **Create Webset**: Organize results into a tracked collection 3. ✨ **Enrich**: Add AI analysis (sentiment, entities, summaries) 4. 📊 **Analyze**: Build graphs to visualize relationships 5. 🔔 **Monitor**: Set up weekly alerts --- ## 📋 Step-by-Step Workflow ### Step 1: Initial Search for Brand Mentions Start by searching for your company name and competitor references across web and news sources. **Tool**: \`web_search_exa\` **Parameters**: \`\`\`json { "query": "${companyName} OR competitors site:news.ycombinator.com OR site:producthunt.com OR site:twitter.com", "numResults": 50, "timeRange": "${period}", "waitForResults": true } \`\`\` **Expected Output**: List of 25-50 results with URLs, titles, and snippets mentioning your company or competitors. --- ### Step 2: Create a Persistent Webset Once you have search results, create a webset to track this data over time. **Tool**: \`websets_manager\` **Parameters**: \`\`\`json { "operation": "create_webset_from_search", "name": "${companyName} - Brand Monitor (${period})", "description": "Tracking brand mentions, competitor activity, and industry sentiment for ${companyName} across news, blogs, and social platforms. Target audience: ${audience}.", "searchQuery": "${companyName} OR competitors site:news.ycombinator.com OR site:producthunt.com OR site:twitter.com", "timeRange": "${period}", "maxResults": 100 } \`\`\` **Expected Output**: \`\`\`json { "websetId": "ws_brand_monitor_123", "status": "processing", "createdAt": "2024-01-10T10:00:00Z", "estimatedCompletion": "2024-01-10T10:30:00Z" } \`\`\` **Note**: Websets are asynchronous. Processing typically takes 10-30 minutes depending on query complexity. --- ### Step 3: Check Webset Status (Optional - if not using webhooks) Poll the webset status until it completes. **Tool**: \`websets_manager\` **Parameters**: \`\`\`json { "operation": "get_webset_status", "websetId": "ws_brand_monitor_123" } \`\`\` **Expected Output**: \`\`\`json { "websetId": "ws_brand_monitor_123", "status": "completed", "itemCount": 87, "completedAt": "2024-01-10T10:25:00Z" } \`\`\` --- ### Step 4: Retrieve and Enrich Items Once completed, fetch the items and add AI enrichments (summaries, entity extraction, sentiment). **Tool**: \`websets_manager\` **Parameters** (Get items): \`\`\`json { "operation": "list_items", "websetId": "ws_brand_monitor_123", "limit": 50, "offset": 0 } \`\`\` **Expected Output**: Array of 50 items with \`url\`, \`title\`, \`text\`, \`publishedDate\`, \`author\`, \`score\`. **Tool**: \`websets_manager\` **Parameters** (Enrich): \`\`\`json { "operation": "enhance_content", "websetId": "ws_brand_monitor_123", "enhancements": [ "summarize", "extract_entities", "classify_sentiment", "identify_topics" ], "waitForResults": true, "advanced": { "outputFormat": "json" } } \`\`\` **Expected Output**: Each item now includes: - \`summary\`: 1-2 sentence summary - \`entities\`: [\`company\`, \`person\`, \`product\`, \`topic\`] - \`sentiment\`: "positive" | "negative" | "neutral" - \`topics\`: ["AI", "Product Launch", ...] --- ### Step 5: Build a Knowledge Graph (Optional - for visualization) Connect entities to understand relationships between companies, people, and topics. **Tool**: \`knowledge_graph\` **Parameters**: \`\`\`json { "operation": "create_entities", "names": ["${companyName}", "Competitor A", "Competitor B"], "entityType": "company" } \`\`\` Then add observations from your webset: **Tool**: \`knowledge_graph\` **Parameters**: \`\`\`json { "operation": "add_observations", "entityName": "${companyName}", "contents": [ "Mentioned in 23 articles this month", "Sentiment: 65% positive", "Top topics: AI, Security, Enterprise", "Key influencers: @alice, @bob" ] } \`\`\` --- ## 💡 Real-World Examples ### Example 1: Share of Voice Analysis **Goal**: Compare how often your company is mentioned vs. top 3 competitors. **Search Query**: \`\`\` ("${companyName}" OR "Competitor A" OR "Competitor B" OR "Competitor C") site:news.ycombinator.com OR site:producthunt.com OR site:techcrunch.com \`\`\` **Enrichment Focus**: Extract company names and count mentions **Expected Insight**: "CompanyA: 34 mentions, CompanyB: 28 mentions, CompanyC: 18 mentions, Competitor1: 12 mentions" --- ### Example 2: Sentiment Tracking Over Time **Goal**: Monitor if brand perception is improving or declining. **Search Query**: \`\`\` "${companyName}" reviews OR feedback OR experience (good OR bad OR amazing OR terrible) \`\`\` **Enrichment Focus**: Classify sentiment, extract topics **Expected Insight**: "Positive sentiment trending up 15% week-over-week. Top complaint: onboarding complexity." --- ### Example 3: Content Gap Discovery **Goal**: Find topics competitors are covering that you should address. **Search Query**: \`\`\` "Competitor A" tutorial OR guide OR how-to -"${companyName}" \`\`\` **Enrichment Focus**: Extract topics, summarize **Expected Insight**: "Competitors publishing 3x more content on 'Machine Learning Ops'. Opportunity to create beginner guides." --- ## 🎓 Best Practices ### Rate Limiting & Performance - Start with 50-100 results to test queries - Scale to 500+ results once query is refined - Use \`waitForResults: false\` for non-blocking creation - Set up webhooks instead of polling for better efficiency ### Enrichment Strategy - Run enrichments in batches (50-100 items per batch) - Prioritize \`summarize\` and \`extract_entities\` first - Add \`classify_sentiment\` for brand health tracking - Use \`identify_topics\` for content gap analysis ### Monitoring Loop - Run brand monitor websets weekly - Compare week-over-week trends - Set alerts for unusual spikes (positive or negative) - Archive completed websets monthly for historical analysis ### Deduplication - Use \`deduplicate_urls\` flag when creating webset to avoid duplicate domains - Tag items by source (news, blog, social) for filtering --- ## ⚠️ Troubleshooting ### Issue: Webset returns 0 or very few results **Cause**: Likely a technical error (API rate limit, network issue, service unavailability). **Solution**: - Check API key is set correctly: \`echo $EXA_API_KEY\` - Verify network connectivity to Exa API - Check recent API status page - Retry the webset creation after waiting a few minutes - If persists, contact support with webset ID and timestamp ### Issue: Enrichment times out **Cause**: Webset too large (1000+ items). **Solution**: - Split into smaller websets (500 items max per enrichment) - Use \`waitForResults: false\` and check status later - Enrich high-priority items first (filter by score) ### Issue: No API key error **Cause**: \`EXA_API_KEY\` environment variable not set. **Solution**: \`\`\`bash export EXA_API_KEY="your-api-key-here" \`\`\` ### Issue: Sentiment classification looks inaccurate **Cause**: Short snippets or ambiguous language. **Solution**: - Request full text extraction in search (\`includeText: true\`) - Manually review 10-20 items and refine query - Use \`classify_intent\` instead for B2B content --- ## ❓ FAQ **Q: How often should I run brand monitoring?** A: Weekly is recommended for active markets. Monthly for stable niches. Adjust based on industry news velocity. **Q: Can I track multiple competitors in one webset?** A: Yes! Use OR logic: \`"Company A" OR "Company B" OR "Company C"\` **Q: How long do websets stay available?** A: Indefinitely—archive and revisit anytime. Exa retains websets for historical analysis. **Q: Can I export the results?** A: Yes! Use \`export_webset\` with \`csv\`, \`json\`, or \`markdown\` format. **Q: What if I want real-time alerts?** A: Set up a webhook to notify when new mentions appear. See \`webhook_setup_guide\` prompt. **Q: Can I combine brand monitoring with other websets?** A: Yes! Use \`knowledge_graph\` to connect separate brand, competitor, and industry websets. --- ## 🚀 Next Steps 1. Customize the company name and target audience for your organization 2. Run your first brand monitor webset 3. Set up a weekly scheduled task to recreate the webset 4. Create enrichment rules to automatically tag and score items 5. Build a dashboard to visualize share of voice and sentiment trends Ready to start? Run the prompt again with your company details, or jump straight to creating your first webset!`; }