Competitors already ran the experiments. They paid for the knowledge. Prowl extracts the outcome.

Connect any agent — Cursor, Claude Code, Codex, or your own stack — to https://prowl.chat/mcp and it stops guessing. It gains SEO & backlinks from two independent indexes, 60+ SERP engines, ad-library intelligence, funnels, reviews, pricing, and AI/LLM mention tracking across ChatGPT / Claude / Gemini / Perplexity.

Prowl doesn't hand your agent a pile of APIs. It plans a research DAG, executes it at zero LLM cost, verifies every claim adversarially, and composes a ready-to-execute strategy report — replacing 10–30 hours of manual research and spreadsheet work.

1. Get a prowl_… API key at prowl.chat → MCP Home → API keys. $5 starter credit, no card.

2. Point your MCP client at https://prowl.chat/mcp (Streamable HTTP):

   {
     "mcpServers": {
       "prowl": {
         "url": "https://prowl.chat/mcp",
         "headers": { "Authorization": "Bearer prowl_YOUR_KEY" }
       }
     }
   }

3. Call prowl_list_tools / prowl_search_tools (free), or run prowl_analyze for a full hunt.

Fastest path — install the plugin. Hand your coding agent the skill and it wires up the client config + token file itself:

claude plugin marketplace add https://github.com/PROWL-AI/prowl-skill
claude plugin install prowl@prowl

Prefer a terminal? npm install -g @prowl-ai/cli → prowl analyze "competitors of stripe.com".

22 MCP tools front the whole catalog (20 logical + 2 legacy aliases) — prowl_analyze, prowl_call_tool, prowl_search_tools, prowl_tool_info, prowl_generate_artifact, prowl_export_report, async research sessions, and recurring schedules.

plan  →  execute DAG  →  extract claims  →  adversarial verify  →  gap-fill  →  compose report
         (zero LLM cost)                    (quorum kill-rule)                 (parallel sections)

• Universal Research Engine — a typed evidence ledger with computed confidence. Claims are extracted, attacked, then refuted or confirmed; contradictions and refuted claims ship inside the report alongside an honest coverage disclosure. Numbers, not vibes.
• Dynamic report shape — a blueprint agent composes sections per query from a 12-section catalog (market topology, positioning matrix, revenue mechanics, funnel intelligence, creative signal, market pulse, voice & vulnerability, SEO growth, market sizing, channel economics, growth signals, idea verdict) — or invents its own.
• 8 playbooks — force a fixed, persona-tuned shape: geo-visibility-audit, competitor-teardown, content-engine, local-and-reputation, mobile-aso, amazon-marketplace, idea-validation, channel-economics-audit.
• 3 tiers — basic / deep / max, each with a hard tool budget and provider-cost cap per run. You are never billed more than the reserved hold.
• Exports — Markdown, PDF, PPTX, audio brief, video.

Pay-as-you-go from a USD wallet. prowl_list_tools, prowl_search_tools, and prowl_tool_info are free; metered calls return a billing object with estimated and actual cost. Insufficient balance blocks a call before it runs — never mid-report.

prowl_… keys are billing-bearing. Treat them as secrets. Never commit one to a public repo.