MCP and agents

Run a multi-agent competitor research sprint

Split competitor research across specialist AI agents for collection, extraction, analysis, verification, and final reporting.

Best use case

Use this prompt when the source set matches the job

Use this with ChatGPT agents, Manus, Claude, Codex, Gemini, Perplexity, or any workflow where AI can run multi-step research.

Before you paste

Give the prompt sources, tools, dates, and a decision

  • Paste raw notes with labels like homepage, pricing page, ad copy, SERP notes, offer page, export, screenshot, or review set.
  • Add the date you checked anything that can change, especially ads, prices, search results, AI answers, and website pages.
  • Tell AI which tools it can use: web search, deep research, files, code, browser, MCP, Semrush, Ahrefs, Similarweb, Panoramata, Sheets, or your own workspace.
  • Tell AI what decision the answer should support, so it gives you a useful recommendation instead of a generic summary.

Modern AI workflow

Use the prompt with current AI tools, not only a blank chat box

  • Use deep research or web search for current public evidence, then cite the URLs and date checked.
  • Use file or data analysis for exports, screenshots, CSVs, and historical logs. Do not summarize rows by instinct.
  • Use MCP/connectors when available so the AI can query Semrush, Ahrefs, Similarweb, Panoramata, Sheets, CRM, or your own files directly.
  • Use agent mode for multi-step research: collect, extract, compare, verify, then write.
  • Use artifacts, Canvas, tables, or charts when the output is a map, report, dashboard, or campaign plan.

Prompt

Run a multi-agent competitor research sprint

You are my competitor intelligence operator.

Task: Plan and execute a multi-agent competitor research sprint with clean handoffs.

My company: {{my_company}}
Competitors: {{competitor}}
Category: {{category}}
Decision I need to support: {{decision}}

Available sources, exports, URLs, files, screenshots, notes, and tool outputs:
{{sources}}

- Map tools, MCP servers, connectors, exports, and manual checks to the exact competitor-intelligence job they should handle.
- Use any provided URLs, files, screenshots, exports, or connected tool outputs before analyzing.
- Cite the source, export, tool, or URL behind any claim that affects the decision.

Do the work:
1. Define the research question and the decision it supports.
2. Create specialist roles: source collector, extractor, analyst, verifier, and report writer.
3. Give each role inputs, outputs, constraints, and stop conditions.
4. Merge findings into one evidence-backed synthesis.
5. Run a verification pass before recommendations.

Return:
- Agent plan.
- Task briefs for each agent.
- Shared evidence table.
- Synthesis report.
- Verification log.

Rules:
- Separate observed evidence, inferred signal, and recommended action.
- Put dates next to any recent or change-based claim.
- Cite URLs when you use web search or deep research.
- Name the tool, connector, MCP server, or export when you use one.
- Do not copy competitor creative. Translate the learning into our company context.
Open in ChatGPT Open in Claude Open in Gemini

Edit the prompt first if needed. ChatGPT and Claude open prefilled; Gemini opens with the prompt copied.

Variables

Replace these fields before you run the prompt

Variable What it means Example
{{my_company}} Required My company The company, product, store, or service you are comparing against the competitor. A DTC skincare brand selling refillable face wash
{{competitor}} Required Competitors One competitor, a short competitor set, or a tracked category. Brand X
{{category}} Required Market or category The buying context. This helps the AI avoid comparing the wrong kind of business. Premium skincare, France and UK
{{sources}} Required Sources and retrieval targets Paste collected sources, exports, screenshots, notes, URLs to check, or the MCP/tool datasets the AI should use. Homepage copy, pricing page, top 5 ads, title tags, Semrush export, Ahrefs export, Similarweb notes, Panoramata campaign examples
{{decision}} Required Decision to support The action you need to take after the analysis. Rewrite our landing page hero and offer comparison table

Example

Use this example to match the right level of detail

Source notes you paste into AI

My company: AI meeting assistant
Competitors: 10 meeting, note, and CRM tools
Sources: URLs, docs, pricing pages, SEO exports, review snippets, campaign archives
Decision: prepare a strategy report for next quarter

What a useful answer should look like

Fictional example output

Agent split:
Collector gets current URLs and exports.
Extractor turns pages into structured tables.
Analyst finds patterns.
Verifier checks unsupported claims.
Writer creates the final report.

Stop condition:
No recommendation without source label and confidence.

Verification

Check whether the answer is useful

  • The output names the evidence behind each recommendation.
  • The output uses current sources, exports, or tool results where the task depends on fresh data.
  • The answer separates facts, estimates, and decisions.
  • The final next moves are specific enough to assign.

Mistakes

Mistakes that make this prompt weak

  • Asking for strategy before the source set is clear.
  • Mixing old screenshots with current claims without dates.
  • Copying competitor language instead of translating the insight.
  • Skipping verification because the answer sounds confident.

Source notes

Use AI to collect data, then make it show the evidence

A good AI workflow can search, inspect pages, analyze exports, call MCP tools, compare screenshots, and build tables. Make it show URLs, dates, exports, screenshots, or connector results behind the answer before you trust the recommendation.

What you should do next

Run it once, then verify the useful parts

Replace the fields, paste a labeled source set, run the prompt, and check the answer before using it in a strategy report.

Use next