Claude MCP server
One endpoint that gives Claude live SERP, ads, traffic and review data instead of recalled guesses.
Prowl is an MCP server for Claude. Add it once and Claude can call 448 market-data tools — live SERPs, keyword and traffic estimates, backlink indexes, ad libraries, review mining — and answer with real numbers. It works in Claude Code and any Claude client that speaks the Model Context Protocol, and it bills per call in USD from a prepaid wallet, so there are no seats and no per-provider contracts.
Add Prowl to Claude Code
One command. Generate a key in MCP Home first, then:
claude mcp add --transport http prowl https://prowl.chat/mcp \
--header "Authorization: Bearer prowl_YOUR_KEY_HERE"Prowl speaks Streamable HTTP at https://prowl.chat/mcp and authenticates with an Authorization: Bearer prowl_<your_key> header, so any client that supports remote MCP servers can use the same endpoint.
To get the /prowl skill alongside the raw tools, install the plugin and restart:
claude plugin marketplace add https://github.com/PROWL-AI/prowl-skill
claude plugin install prowl@prowlAdd Prowl to a JSON-configured Claude client
Clients that take an mcpServers block — Claude Desktop among them — use the same endpoint and header:
{
"mcpServers": {
"prowl": {
"url": "https://prowl.chat/mcp",
"headers": {
"Authorization": "Bearer prowl_YOUR_KEY_HERE"
}
}
}
}Reload the client afterwards. The full first-run path — account, key, wallet top-up, verification and your first report — is on the Getting Started walkthrough.
What Claude can do once it is connected
Claude discovers the catalog with prowl_search_tools, reads a schema with prowl_tool_info, and runs anything with prowl_call_tool. For a full multi-agent report it calls prowl_analyze.
- Live SERPs — rankings, SERP features and competitors for a keyword set, across 60+ engines.
- Backlinks from two independent indexes — DataForSEO and Majestic, so a claim can be cross-checked rather than trusted.
- Demand data — search volume, keyword difficulty, CPC and clickstream estimates.
- Ads and reviews — competitor ad creatives, review mining, listing and GMB data.
- AI-surface visibility — how often a brand is mentioned across ChatGPT, Claude, Gemini and Perplexity.
What it costs
Every tool call debits provider cost times a uniform markup from a prepaid USD wallet, and the response carries a billing object with the exact amount charged, so spend is auditable per call rather than per month. New accounts start with $5 in credit and the basic tier runs without a subscription.
Per-call costs are typically cents: a domain's ranked keywords runs about $0.04, and sizing sixteen competitors' backlink profiles in a single call about $0.07. Billing and top-ups covers the wallet in full, and what to expect from a report run covers per-tier caps before you spend anything.
Frequently asked questions
Does Prowl work with Claude Code and Claude Desktop?
Yes. Prowl is a remote MCP server over Streamable HTTP, so any Claude client that supports remote MCP servers can connect to the same endpoint with the same Bearer header. Claude Code adds it with one `claude mcp add` command; JSON-configured clients use an mcpServers block.
Do I need a subscription to use the Claude MCP server?
No. Every new account gets $5 in starter credit and the basic tier runs unsubscribed. The deep and max report tiers require an active plan; request one without a subscription and the run is downgraded to basic rather than refused.
How is this different from Claude's built-in web search?
Web search returns pages. Prowl returns structured metrics from commercial data providers — ranking positions, search volume, keyword difficulty, backlink counts from two independent indexes, ad creatives, review data. Claude can cite a number and say where it came from instead of summarising an article about it.
Is my API key sent to the tool providers?
No. Your Prowl key authenticates you to Prowl only. Prowl holds the provider credentials and meters the calls, which is why there are no per-provider contracts to sign. Keys support optional restrictions: allowed tool categories, daily and total spend caps, and an IP allowlist.