Competitive Digital Intelligence MCP Server

Competitive digital intelligence for AI agents — this MCP server delivers a full-spectrum analysis of any competitor's digital presence by orchestrating 8 specialist data sources in parallel. It is built for strategy teams, product managers, and AI-powered research workflows that need structured, scored intelligence rather than raw scraped data.

Every tool call returns not just raw data but a scored result: a 0-100 Competitive Intelligence Score built from four weighted dimensions covering technology sophistication, search dominance, e-commerce strength, and brand reputation. Feed the output directly into your AI agent's context window and ask it to generate strategy recommendations.

What data can you extract?

Why use Competitive Digital Intelligence MCP?

Building a competitive analysis by hand means opening 8 different tools, exporting 8 different CSVs, and spending half a day normalizing data before you can say anything meaningful. By the time the slide deck is done, the data is stale.

This MCP server automates the entire pipeline. A single tool call fires all 8 data sources in parallel, scores each dimension against a calibrated 0-100 model, and returns a structured JSON report with verdict and strategic recommendations — in under two minutes.

• Scheduling — run weekly competitive sweeps automatically to track score changes over time
• API access — call from any AI agent, Python workflow, or HTTP client with one request
• Parallel execution — all 8 data sources run simultaneously via Promise.all, not sequentially
• MCP-native — works with Claude Desktop, Cursor, Windsurf, Cline, and any MCP-compatible client
• Spending controls — each tool enforces per-event charge limits so your agent cannot overspend

Features

• 8-source parallel orchestration — fires website-tech-stack-detector, serp-rank-tracker, shopify-store-intelligence, ecommerce-price-monitor, multi-review-analyzer, trustpilot-review-analyzer, wayback-machine-search, and wcag-accessibility-auditor simultaneously via Promise.all
• Tech stack sophistication scoring — compares detected technologies against a curated MODERN list (React, Next.js, Vue, Nuxt, Svelte, Angular, TypeScript, GraphQL, Tailwind, Vercel, Cloudflare, AWS, GCP, Stripe) and LEGACY list (jQuery, Flash, ASP.NET, ColdFusion, Dreamweaver, FrontPage) to produce a score from 0-100
• Four-tier score breakdown — each analysis returns a sub-score (tech stack 0-100, SEO 0-100, e-commerce 0-100, reputation 0-100) plus a composite weighted score
• Composite weighting formula — SEO 30% + e-commerce 25% + reputation 25% + tech stack 20%, calibrated to reflect market impact
• Five-level verdict labels — WEAK_COMPETITOR, EMERGING, ESTABLISHED, STRONG, MARKET_LEADER for instant executive communication
• WCAG-as-SEO signal — accessibility violations (critical/serious) are factored into the SEO score because Core Web Vitals and accessibility correlate with search rankings
• Wayback evolution scoring — snapshot count (20+, 50+ thresholds) contributes to tech stack history score, capturing web longevity and investment continuity
• Actionable recommendations — five conditional recommendation strings generated automatically based on detected patterns (legacy tech, dominant SERP, large catalog, excellent reputation)
• Granular price competitiveness signal — counts discounted SKUs from the price monitor and flags aggressive pricing strategies when 5+ products carry discounts
• Standalone focused tools — six targeted tools (tech_stack_analysis, seo_ranking_intelligence, ecommerce_competitive_analysis, reputation_monitor, price_intelligence, website_evolution_tracker) for partial analyses at lower cost
• Spending limit enforcement — every tool checks chargeResult.eventChargeLimitReached before executing and returns a structured error if the limit is reached, protecting agent budgets

Use cases for competitive digital intelligence

Marketing competitive benchmarking

Strategy teams and CMOs need a defensible number when presenting competitive positioning to leadership. Run full_competitive_audit monthly against your top 3-5 competitors and track the composite score over time. When a competitor's score jumps from 48 to 67 in a quarter, you know they made a significant investment and can investigate where.

Product team technology intelligence

Product managers and engineering leads use tech_stack_analysis to detect when competitors adopt new frameworks or infrastructure. If a competitor migrates from jQuery to React with a Cloudflare CDN layer, that is a signal of technical investment that may precede a UX overhaul or performance push. The Wayback Machine dimension reveals when the migration happened.

E-commerce pricing strategy

Retail analysts and category managers use ecommerce_competitive_analysis and price_intelligence to monitor competitor product breadth and promotional patterns. Knowing that a competitor has 347 products and 28 actively discounted SKUs — versus your 120 products with 4 discounts — informs both assortment and promotional planning.

Investor and M&A digital due diligence

Investment analysts use full_competitive_audit and tech_stack_analysis during target assessment. A LEGACY tech stack verdict with 3+ legacy indicators is a proxy for engineering debt. A Trustpilot score below 3.0 alongside a DOMINANT competitor creates a clear narrative around market risk. The structured JSON output feeds directly into investment memo templates.

Reputation risk monitoring

Brand managers and customer experience leads use reputation_monitor to track review sentiment across Trustpilot and multi-platform sources simultaneously. The scoring model flags when a brand has 5+ negative reviews and when the Trustpilot average diverges significantly from multi-platform data — an early warning signal for reputational divergence.

AI agent competitive research workflows

AI engineers building autonomous research agents embed this MCP to give their agents structured competitive context. An agent tasked with "analyze the top 3 competitors in the CRM market" can call compare_competitors for each one, receive a normalized score with signals, and synthesize a comparative brief without writing any scraping code.

How to connect this MCP server

Claude Desktop

Add this to your claude_desktop_config.json:

{
  "mcpServers": {
    "competitive-digital-intelligence": {
      "url": "https://competitive-digital-intelligence-mcp.apify.actor/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_APIFY_TOKEN"
      }
    }
  }
}

Cursor / Windsurf / Cline

Add the MCP server URL in your IDE's MCP configuration panel:

URL: https://competitive-digital-intelligence-mcp.apify.actor/mcp
Auth: Bearer YOUR_APIFY_TOKEN

Programmatic HTTP (cURL)

# Call full_competitive_audit
curl -X POST "https://competitive-digital-intelligence-mcp.apify.actor/mcp" \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer YOUR_APIFY_TOKEN" \
  -d '{
    "jsonrpc": "2.0",
    "method": "tools/call",
    "params": {
      "name": "full_competitive_audit",
      "arguments": {
        "competitor": "hubspot.com",
        "industry": "CRM"
      }
    },
    "id": 1
  }'

MCP tools reference

Tool parameters

Output example

full_competitive_audit response for acmecorp.com:

{
  "competitor": "acmecorp.com",
  "compositeScore": 68,
  "verdict": "STRONG",
  "techStack": {
    "score": 72,
    "techsDetected": 14,
    "modernStack": 7,
    "legacyIndicators": 1,
    "stackLevel": "ADVANCED",
    "signals": [
      "7 modern technologies — cutting-edge stack",
      "1 legacy technology — technical debt indicator"
    ]
  },
  "seoCompetitive": {
    "score": 65,
    "rankedKeywords": 18,
    "topPositions": 6,
    "seoLevel": "STRONG",
    "signals": [
      "4 keywords in top 3 — dominant SERP position",
      "6 keywords in top 10 — strong organic visibility"
    ]
  },
  "ecommerce": {
    "score": 58,
    "productCount": 247,
    "priceCompetitiveness": 9,
    "ecomLevel": "ESTABLISHED",
    "signals": [
      "247 products — extensive catalog",
      "9 discounted products — aggressive pricing strategy"
    ]
  },
  "reputation": {
    "score": 71,
    "avgRating": 4.2,
    "reviewCount": 142,
    "reputationLevel": "GOOD",
    "signals": [
      "Trustpilot 4.3/5 — strong customer satisfaction"
    ]
  },
  "allSignals": [
    "7 modern technologies — cutting-edge stack",
    "1 legacy technology — technical debt indicator",
    "4 keywords in top 3 — dominant SERP position",
    "6 keywords in top 10 — strong organic visibility",
    "247 products — extensive catalog",
    "9 discounted products — aggressive pricing strategy",
    "Trustpilot 4.3/5 — strong customer satisfaction"
  ],
  "recommendations": [
    "Competitor has cutting-edge tech stack — invest in technical parity",
    "Large product catalog — consider niche specialization to compete",
    "Excellent reputation — differentiate on unique value props"
  ]
}

Output fields

Competitive Intelligence Score explained

The composite score is a weighted average of four dimension scores:

Composite scoring formula: composite = (seo × 0.30) + (ecommerce × 0.25) + (reputation × 0.25) + (techStack × 0.20)

Verdict thresholds:

How much does it cost to run competitive digital intelligence?

Each tool call is priced at $0.045 per call regardless of which tool you use. All 8 tools share the same per-event price. Platform compute costs are included.

You can set a maximum spending limit per session in your MCP client to control costs. The actor enforces this limit per-event and returns a structured error when reached rather than silently failing.

Compare this to SimilarWeb Pro at $167/month or Semrush at $120/month — most users running regular competitive sweeps with this MCP spend under $5/month with no subscription commitment. Apify's free tier includes $5 of monthly credits, making most light use cases free.

How Competitive Digital Intelligence MCP works

Phase 1: Parallel data collection

When a tool call arrives, runActorsParallel fires all required sub-actors simultaneously using Promise.all. Each sub-actor is called via its Apify actor ID (hardcoded in ACTOR_MAP) with 512 MB memory and a 120-second timeout. Results that fail or time out return empty arrays, so the scoring model always receives a complete (if sparse) data structure.

The 8 sub-actors cover distinct signal types: mBdMOhOMY7XtwX99T (tech stack), FdqNJLv7xJSFTuuYe (SERP), eiHLUkf3SgN0tv0JL (Shopify), hgGm1sLFsSfwCqqiH (price monitor), S85IfVOoTyN9XWyXs (multi-review), BLXRNH33QZNOoyKEg (Trustpilot), rT8Qt6fe3ygVyVMdb (Wayback), IZc7S8V54w3JfvnUU (WCAG).

Phase 2: Four-dimension scoring

scoreTechStack scores tech sophistication across four sub-components: tech sophistication (max 40, based on MODERN/LEGACY list matching with weights of +5 per modern tech and -3 per legacy tech), Wayback evolution (max 25, based on snapshot count thresholds), tech diversity (max 20, based on unique technology count), and stack maturity (max 15, based on modern-vs-legacy ratio and breadth). All sub-components are clamped and summed to produce the 0-100 score.

scoreSEOCompetitive weights SERP positions as: top-3 keywords worth 8 points each, top-10 worth 4 points each, top-20 worth 2 points each (max 40). WCAG violations contribute an accessibility sub-score (25 points for zero critical/serious violations, 15 for 3 or fewer, 8 for 10 or fewer). Keyword breadth and SEO maturity add the remaining 35 points.

scoreEcommerce derives product count from Shopify data and discount count from the price monitor, then applies tiered thresholds: 500+ products scores 35/40, 100+ scores 25/40, 20+ scores 15/40. Price competitiveness scores each discounted product at 4 points (max 30 combined with price breadth).

scoreReputation multiplies the average multi-platform rating by 8 (max 40 from reviews) and the average Trustpilot rating by 6 plus review volume (max 30 from Trustpilot). Volume and consistency bonuses add the remaining 30 points.

Phase 3: Composite assembly and recommendations

generateCompetitiveIntel applies the weighted formula, assigns a verdict label, collects all signals from all four scorers into a flat allSignals array, and generates up to 5 conditional recommendations based on detected patterns (cutting-edge stack, dominant SERP, large catalog, excellent or poor reputation, legacy tech opportunity).

Phase 4: MCP transport

The server runs in Apify Standby mode on port ACTOR_STANDBY_PORT (default 8080) using StreamableHTTPServerTransport from the @modelcontextprotocol/sdk. Each POST to /mcp creates a fresh McpServer instance bound to a new transport, processes the request, and closes the transport. When not in standby mode, the server starts briefly, logs a health check message, and exits after 1 second.

Tips for best results

1. Use full_competitive_audit for new competitors. The first time you analyze a competitor, the full audit gives you all four dimension scores at once for the same $0.045 as any other tool. Use focused tools only for ongoing monitoring when you already know which dimension matters.

2. Pass the industry parameter for better recommendations. The industry context is logged but also helps your AI agent interpret the recommendations in domain context. For SaaS CRM competitors, "large product catalog" means something different than for apparel e-commerce.

3. Schedule weekly compare_competitors calls for your top 3 rivals. At $0.045/call, tracking 3 competitors weekly costs $0.54/month. Store the scores in a dataset and use Website Change Monitor to flag when scores change significantly.

4. Combine with company-deep-research for investment research. The competitive digital score tells you about digital presence; company deep research tells you about financials, leadership, and public information. Together they cover the full due diligence surface.

5. For e-commerce-heavy competitors, run ecommerce_competitive_analysis and price_intelligence separately. The dedicated tools return more pricing rows (up to 25 for price_intelligence vs 20 in the full audit) and more Shopify data rows (15 vs the full audit's truncation).

6. Interpret LEGACY tech stack results carefully. Legacy indicators like jQuery do not always mean poor engineering — many high-performing sites use jQuery for specific components. The signal is most meaningful when legacyIndicators >= 2 alongside a BASIC or LEGACY stackLevel verdict.

7. Use website_evolution_tracker before pitch meetings. Knowing that a competitor's Wayback snapshot count jumped from 30 to 89 in 18 months tells a story about growth momentum that the current-state audit misses.

Combine with other Apify actors

Limitations

• Shopify-only e-commerce data — the e-commerce dimension covers Shopify stores only. WooCommerce, Magento, custom storefronts, and Amazon seller accounts are not detected. Competitors on non-Shopify platforms will receive low e-commerce scores regardless of actual catalog size.
• SERP rankings reflect tracked keywords only — the SEO score is proportional to what the SERP Rank Tracker finds for the input query. Competitors with strong positions on long-tail keywords not covered by the query will be underscored. For comprehensive keyword gap analysis, use a dedicated SEO platform.
• No JavaScript rendering in tech detection — the tech stack detector analyzes HTTP headers, static HTML, and resource URLs. Single-page applications that load frameworks dynamically after the initial page render may have technologies go undetected.
• Trustpilot requires a listing — brands without a Trustpilot presence return empty arrays for the Trustpilot dimension, reducing the maximum achievable reputation score. The score normalizes for this but the dimension will be weaker.
• Wayback Machine coverage varies by domain age — newer domains (under 3 years) have few snapshots and will score low on the Wayback evolution sub-component regardless of their actual investment level. Do not over-index on tech stack scores for recently launched competitors.
• Data freshness — all data is fetched live at call time. There is no caching layer. Scores for the same competitor may vary slightly between runs if the underlying platforms have been updated.
• Sub-actor timeouts — each sub-actor is allowed 120 seconds. If a source returns slowly or fails, it contributes an empty array. The composite score will be lower but not invalid. High-timeout scenarios are most common for large Shopify stores.
• No mobile app intelligence — this server covers web properties only. Competitors whose primary channel is a mobile app will score lower than their actual digital presence warrants.

Integrations

• Zapier — trigger full_competitive_audit weekly and push composite scores to a Google Sheet for trend tracking
• Make — build competitive monitoring workflows that alert Slack when a competitor's score increases by more than 10 points
• Google Sheets — export scored competitor data for executive dashboards and strategy presentations
• Apify API — integrate into Python or JavaScript data pipelines for automated competitive research
• Webhooks — post allSignals and recommendations to your internal Slack channel or CRM after each run
• LangChain / LlamaIndex — use structured competitive intel JSON as retrieval context for LLM-powered strategy agents

Troubleshooting

Composite score is unexpectedly low despite a well-known competitor. The most common cause is the competitor using a non-Shopify storefront (scores 0 on e-commerce), lacking a Trustpilot listing (partial reputation score), or the SERP query not matching the keywords they rank for. Check ecomLevel, reputationLevel, and rankedKeywords individually to identify which dimension is dragging the composite score. Try passing a more specific keyword string in seo_ranking_intelligence to improve SERP coverage.

Run takes longer than 2 minutes. The server allows 120 seconds per sub-actor. If several sub-actors run slowly simultaneously, total wall time can approach 3-4 minutes for full_competitive_audit. This is normal for large Shopify stores with thousands of products or domains with many Wayback snapshots. If you need faster results, use a focused tool (e.g., tech_stack_analysis or reputation_monitor) that calls only 2 sub-actors.

All e-commerce fields return 0. The competitor's storefront is likely not built on Shopify, or Shopify Intelligence did not find a matching store for the input query. Try the competitor's exact domain rather than brand name. If the competitor uses WooCommerce or a custom platform, the e-commerce dimension will not produce meaningful data with this MCP.

error: Spending limit reached in tool response. Your MCP session has hit the configured per-event spending limit. Either increase the limit in your Apify account settings or your AI agent's session configuration. This is a protection mechanism — each tool checks chargeResult.eventChargeLimitReached before executing.

Tech stack signals show LEGACY for a clearly modern brand. The detection model matches technology names from the website-tech-stack-detector output against hardcoded keyword lists. If the detected technology names use vendor-specific terminology not covered by the MODERN list (e.g., a proprietary CDN or framework), the modern count will be lower than expected. The raw technologies array in the response shows exactly what was detected for manual review.

Responsible use

• This server only accesses publicly available website data, search results, review platforms, and the Internet Archive.
• Respect website terms of service and robots.txt directives.
• Competitive intelligence for legitimate business strategy and research is the intended use case.
• Do not use extracted data to facilitate unauthorized access, defamation, or targeted harassment of brands or individuals.
• Review sentiment and reputation scores are derived from publicly posted customer reviews and reflect real customer experiences.
• For guidance on web scraping legality, see Apify's guide.

FAQ

How is Competitive Digital Intelligence MCP different from SimilarWeb or Semrush? SimilarWeb and Semrush use proprietary traffic panel data and large keyword databases built over years. This MCP fetches live public data at call time — SERP results, real review counts, actual Wayback snapshots — and scores them through a transparent, auditable scoring model. The output is structured JSON designed for AI agent consumption rather than a human dashboard. It costs $0.045/call vs $120-167/month for those platforms.

What does the competitive intelligence score actually measure? Four dimensions: technology sophistication (20%), search engine ranking strength (30%), e-commerce presence (25%), and brand reputation (25%). Each dimension has its own 0-100 score with a named level label. The composite is a weighted average. See the scoring formula section for exact weights and component breakdowns.

How many data sources does a full_competitive_audit call? Eight: website-tech-stack-detector, serp-rank-tracker, shopify-store-intelligence, ecommerce-price-monitor, multi-review-analyzer, trustpilot-review-analyzer, wayback-machine-search, and wcag-accessibility-auditor. All run in parallel via Promise.all, not sequentially.

Can I run competitive intelligence on non-English websites? Yes. The tech stack detector, Wayback Machine, and WCAG auditor work on any domain regardless of language. The SERP tracker results depend on the keywords you provide. Review platforms may return fewer results for brands primarily reviewed in non-English markets.

How accurate is the SERP ranking data? SERP data reflects actual search results at the time of the call for the keywords derived from your input query. It is not a historical database — it shows current rankings for the query term. For tracking keyword positions over time, call seo_ranking_intelligence on a schedule and store the results.

Does competitive digital intelligence work for local businesses? Yes, with caveats. Local businesses typically have small or no Shopify presence (low e-commerce score), few Wayback snapshots (lower tech evolution score), and limited SERP visibility on non-local keywords. The reputation dimension tends to be the most useful for local businesses given their Trustpilot and review platform presence.

How long does a typical full_competitive_audit call take? Between 60 and 120 seconds for most competitors, since all 8 sub-actors run in parallel. The slowest sub-actor determines total wall time. Shopify stores with large catalogs and heavily archived domains tend to be the slowest.

Is it legal to scrape competitor data for competitive intelligence? Generally yes for publicly available data. Web scraping publicly accessible information for competitive research is a well-established and legally recognized practice. The specific legal landscape varies by jurisdiction. See Apify's legal guide for detailed guidance.

Can I compare my own domain against a competitor using this MCP? Use compare_competitors and pass your domain in the benchmark parameter. The tool accepts both inputs and returns a competitive summary including both domains' scores. For a full bilateral analysis, run full_competitive_audit on both domains and compare the JSON output.

What happens if one of the 8 sub-actors fails or times out? The failing sub-actor returns an empty array. The scoring model handles missing data gracefully — missing dimensions produce lower scores for that component but do not break the composite calculation. The allSignals array will be smaller and some score components will be near zero rather than reflecting no data.

Can I schedule this to run weekly competitive sweeps automatically? Yes. Use Apify Schedules to configure any tool as a recurring run. You can set the schedule in the Apify console, via the API, or through Zapier/Make. Results are stored in the Apify Dataset for that run and can be exported as JSON, CSV, or pushed to external systems via webhooks.

How do I interpret a WEAK_COMPETITOR verdict? It means the competitor scores below 20/100 on the composite model. The most common causes are: no Shopify presence (reducing e-commerce score), no Trustpilot listing (reducing reputation score), and low SERP visibility for the query used. It may indicate a genuinely weak competitor, or it may indicate the competitor operates on channels not covered by this MCP (Amazon, app stores, direct sales). Always inspect the individual dimension scores before drawing conclusions.

Help us improve

If you encounter issues, help us debug faster by enabling run sharing:

1. Go to Account Settings > Privacy
2. Enable Share runs with public Actor creators

This lets us see your run details when something goes wrong. Your data is only visible to the actor developer, not publicly.

Support

Found a bug or have a feature request? Open an issue in the Issues tab on this actor's page. For custom competitive intelligence workflows, AI agent integrations, or enterprise use cases, reach out through the Apify platform.