Guides · For AI agents
Is there an MCP server for competitive marketing intelligence?
Last updated November 11, 2026Servers checked November 11, 202610 serversBy Every Signal
Yes. Every Signal runs a remote MCP server at https://everysignal.com/mcp. An agent connected to it can order an automated snapshot of a competitor (tracking stack, live Meta ads, search footprint, homepage), score a brand's live ads, audit a domain's growth and search footprint, and read a knowledge base of real funnels, copy and measured conversion rates from teardowns. Each tool is paid per call from prepaid credits, with 50 free credits on your first key.
It is not the only MCP server a marketer needs. Below is the exact config, what each Every Signal tool returns and costs, prompts that work, and a verified list of nine other MCP servers for marketing work, from Google Analytics and Google Ads to Ahrefs, Semrush, Similarweb, DataForSEO, Apify, Firecrawl and a community Meta Ads server.
The tools
What does the Every Signal MCP server do?
Tools, credits and prices as listed on /connect/ on November 11, 2026. 1 credit is $0.10. Every job debits up front and refunds itself if it fails.
| Tool | What it returns | Credits | Price |
|---|---|---|---|
| intel_report | An automated snapshot of one company: tracking stack, live Meta ads, search footprint, homepage, and a machine summary against its sector. Runs 6 to 12 minutes and returns a job id. No analyst review. | 490 | $49 |
| analyze_ads | Up to 10 ad creatives (image URLs or ad text), or a brand's live Meta ads, scored against real ad patterns in the knowledge base. Fixes per creative and new hooks. | 150 | $15 |
| growth_audit | What a domain ranks for, its search and AI-answer footprint and tracking, compared with the growth moves of real companies in its sector. A ranked growth plan. | 150 | $15 |
| analyze_landing_page | One public URL: tech stack and pixels, funnel steps, copy against sector patterns, speed signals. Ranked fixes and a hero rewrite. | 100 | $10 |
| audit_work | Copy, a landing page, an email, an ad, a funnel or a paywall your AI wrote, plus the sector. A 0 to 100 score, findings with the record ids behind them, fixes and a rewrite. About 40 seconds. | 30 | $3 |
| kb_search, kb_examples, kb_benchmarks, kb_get | Search the knowledge base, pull real copy by sector and section, read measured conversion rates with their evidence tags, fetch one record. | 1 each | $0.10 |
| kb_sectors, kb_brief, kb_playbook | List sectors, read a sector brief, read a playbook for a task (homepage, email, paid social, subscription funnel, checkout, GEO). | 0 | free |
| list_reports, get_report, job_status, credits_balance | List published teardowns, read one (full text with library access, else the public summary), poll a job, read your balance. | 0 | free |
Behind the tools is a knowledge base built from Every Signal's own teardowns: 52 companies across 16 public reports on the day of writing verified, plus the records they produced. Every figure in a record carries an evidence tag: verified, estimate, inferred or rumour. Nothing is read from behind a login; pages are fetched by EverySignalBot. Limits are 60 calls a minute and 5 running jobs per key.
Setup
How do I connect Every Signal to Claude, Cursor or ChatGPT?
Get a key at /app/keys (sign in with an email code; keys start with es_). Then paste one block. The server uses Streamable HTTP with a Bearer header. The configs below are copied from /connect/, which also has Windsurf and curl.
Claude Code
terminal
claude mcp add --transport http everysignal https://everysignal.com/mcp --header "Authorization: Bearer YOUR_KEY"
Cursor, and any client that takes a URL plus headers
~/.cursor/mcp.json
{"mcpServers":{"everysignal":{"url":"https://everysignal.com/mcp","headers":{"Authorization":"Bearer YOUR_KEY"}}}}Claude.ai and ChatGPT, no key to paste
In Claude: Settings, Connectors, Add custom connector, paste the URL below and click Connect. In ChatGPT with developer mode on: Settings, Apps and Connectors, Advanced, create a connector with the same URL and OAuth. You sign in to Every Signal by email; the first connection comes with 50 free credits.
connector url
https://everysignal.com/mcp/connect
Claude Desktop
Claude Desktop talks to local servers, so it reaches a remote one through mcp-remote (needs Node). Add this to claude_desktop_config.json and restart the app.
claude_desktop_config.json
{"mcpServers":{"everysignal":{"command":"npx","args":["-y","mcp-remote","https://everysignal.com/mcp","--header","Authorization: Bearer YOUR_KEY"]}}}The server is also listed in the official MCP Registry as com.everysignal/everysignal, and there is a Claude skill that teaches an agent when to call which tool.
Prompts
What should I ask it?
The agent picks the tool from the prompt. These phrasings map cleanly to one tool each, and the cost is what that tool charges.
| Prompt | Tool it calls | Cost |
|---|---|---|
Build an intelligence report on example.com and tell me what its tracking stack says about where it spends. | intel_report | $49 |
Pull this brand's live Meta ads, score them, and give me three hooks to test against them. | analyze_ads | $15 |
What does example.com rank for, and what would a ranked growth plan look like against its sector? | growth_audit | $15 |
Analyze this landing page URL against how subscription apps sell and rewrite the hero. | analyze_landing_page | $10 |
Audit the trial email you just drafted for a subscription app before I send it. | audit_work | $3 |
What measured trial-to-paid or quiz completion rates does the knowledge base have for subscription apps, with evidence tags? | kb_benchmarks | $0.10 |
Show me real paywall copy from meditation apps. | kb_examples | $0.10 |
Which sectors does the knowledge base cover, and what is the paid social playbook? | kb_sectors, kb_playbook | free |
Two habits make the output more useful. Name the sector when you know it, because the audit and analysis tools compare against sector records. And ask the agent to keep the evidence tags in its summary, so an estimate does not come back as a fact.
The rest of the stack
Which other MCP servers are useful for marketing work?
Each entry was checked against the publisher's own repository or docs on November 11, 2026. Official
means the vendor publishes it. Pricing is the vendor's; most need that vendor's paid plan or API key.
| Server | What it covers | Access |
|---|---|---|
| Google Analytics MCP official, Google | Your own GA4 properties: account summaries, reports, funnel and conversion reports, realtime data. Write tools are opt-in. stdio or HTTP. | Google Cloud credentials for a GA user; no paid plan |
| Google Ads MCP official, Google | Your own Google Ads accounts through GAQL queries, resource metadata and a propose-then-confirm flow for writes. Not competitors' ads. | Google Ads API access with OAuth |
| Ahrefs MCP official, hosted | Ahrefs' SEO data: backlinks, keywords, organic traffic estimates, through a remote server at api.ahrefs.com/mcp/mcp. The older local repo is archived. | Paid plan, Lite or higher; uses API units |
| Semrush MCP official, hosted | Semrush's SEO API data, the Trends API (traffic and market data) with a Trends subscription, and read-only Projects data. mcp.semrush.com/v2/mcp, Streamable HTTP only. | Paid plan with API units |
| Similarweb MCP official, hosted | 24 tools: traffic and engagement, traffic sources, rank, similar sites, referrals, audience overlap, popular pages, PPC spend, ad networks, geography, technologies, demographics, two search tools and three app tools. mcp.similarweb.com. | Subscription with API access and an API key; uses data credits |
| DataForSEO MCP official | Documentation tools plus an authenticated api_request tool for any DataForSEO endpoint: SERPs, backlinks, keywords, and more. stdio, local HTTP or hosted at mcp.dataforseo.com/v3/mcp. | DataForSEO API credentials, pay as you go |
| Apify MCP official | Search the Apify Store, run Actors (scrapers, including ad-library and social scrapers published there) and read their datasets. Hosted at mcp.apify.com or local stdio. | Connecting is free; Actor runs bill to your Apify account |
| Firecrawl MCP official | Scrape pages to Markdown, map and crawl sites, search, parse PDFs, monitor pages, enrich companies. 31 tools. stdio or hosted at mcp.firecrawl.dev/v2/mcp. | Keyless free tier for scrape, search and parse; API key for the rest |
| Meta Ads MCP community, Pipeboard; not Meta | Your own Meta ad accounts: campaigns, ad sets, ads, creatives, insights, targeting search. 42 tools. Its README does not cover the Ad Library, so it does not read competitors' ads. | Hosted via Pipeboard login, or self-host with your own Meta developer app |
To find more, use the official MCP Registry, a searchable directory run by the protocol's maintainers. The reference servers repository holds the protocol's own example servers (fetch, filesystem, memory and so on) and points to the registry for third-party servers. The protocol itself is documented at modelcontextprotocol.io.
Putting it together
How do these servers fit together?
Your own numbers: Google Analytics, Google Ads, Meta Ads
These servers read accounts you own. They answer how are we doing
, not what is the competitor doing
. They are the cheapest to run because the data is already yours.
Size and search: Similarweb, Semrush, Ahrefs, DataForSEO
These estimate how much traffic a site gets and what it ranks for, across almost any domain. Their MCP servers sit behind paid plans and meter API units or credits. A startup that does not already pay for one of them will not get far with the MCP server alone.
Mechanism: Every Signal
Every Signal answers why a competitor is growing and what a customer sees: the live ads and where they land, the funnel and paywall, the tracking stack, the pricing, with every figure tagged. It has no traffic panel of its own and cites third-party estimates as estimates. For the split between it and a traffic tool, read Every Signal vs Similarweb.
Raw pages: Firecrawl, Apify
When no server has the data, these fetch public pages and run scrapers, and your agent does the reading. That is more work per answer but covers anything public.
Questions
FAQ
What is an MCP server, in one paragraph?
MCP, the Model Context Protocol, is an open standard for connecting AI applications to outside systems. An MCP server publishes a list of tools; a client such as Claude, Cursor or ChatGPT reads the list and lets the model call the tools during a conversation. Every Signal's server publishes its audit, analysis, report and knowledge-base tools that way, so an agent can order a competitor snapshot or read measured conversion rates without custom code.
Does the Every Signal MCP server cost anything to connect?
No. Connecting is free and the first API key on a verified email comes with 50 free credits. Tools are then paid per call from prepaid credits at $0.10 a credit: knowledge-base lookups cost 1 credit, an audit 30, a landing page analysis 100, an ad or growth analysis 150 and an intelligence report 490. Sector briefs, playbooks, job status and balance are free. Failed jobs refund themselves.
Can an MCP server show me a competitor's ads?
Every Signal's analyze_ads and intel_report tools read a brand's live Meta ads from Meta's public Ad Library and score them against real ad patterns. Similarweb's MCP server reports PPC spend estimates and ad networks for a domain. The Google Ads and Meta Ads MCP servers listed here read your own ad accounts, not a competitor's, which is a common point of confusion.
Which MCP server should a startup connect first?
Connect the servers that match the data you already pay for. Google Analytics is free and reads your own numbers. Firecrawl and Apify have free tiers for fetching public pages. Ahrefs, Semrush and Similarweb require a paid plan with API access before their MCP servers work. Every Signal is prepaid per call with no subscription, which suits a team that asks a few competitive questions a week.
Can I use Every Signal from ChatGPT or a custom GPT?
Yes. With developer mode on, ChatGPT takes the connector URL https://everysignal.com/mcp/connect with OAuth sign-in by email. Without developer mode, a custom GPT can import the OpenAPI spec at https://everysignal.com/api/v1/openapi.json as an Action with a Bearer API key.
Connect it
Give your agent a competitive intelligence tool today
One key, one config block, 50 free credits. The first free credits cover one audit and a handful of knowledge-base lookups; an intelligence report on a competitor is $49 in credits with no subscription. Setup for every client is at /connect/; packs and prices are on /pricing/.