Connect Prowl to your agent in 5 steps

1 — Create your account

Sign up at prowl.chat with email and password or with Google. New accounts receive $5 in starter credit, which is enough for light testing on the basic tier.

2 — Generate an API key

In the web app open MCP Home → Set up in 3 steps → Step 1 “Generate your token” and click Generate token. The key looks like prowl_ followed by 32 hex characters.

It is shown exactly once — the panel warns “This key will not be shown again.” Copy it somewhere safe before closing it. Lost a key? Generate a new one and revoke the old.

Optional hardening at creation time: allowed and denied tool categories, a daily spend limit, a total spend limit, and an IP allowlist.

3 — Connect your client

Prowl speaks Streamable HTTP at https://prowl.chat/mcp and authenticates with an Authorization: Bearer prowl_<your_key> header. Pick your client:

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

claude mcp add --transport http prowl https://prowl.chat/mcp \
  --header "Authorization: Bearer prowl_YOUR_KEY_HERE"

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

[mcp_servers.prowl]
url = "https://prowl.chat/mcp"

[mcp_servers.prowl.headers]
Authorization = "Bearer prowl_YOUR_KEY_HERE"

Prefer to let the agent wire itself up? Give it this prompt:

Set up the Prowl MCP for this project using the install skill at
https://prowl.chat/mcp/install.md — my token is prowl_YOUR_KEY.
Write the client config plus a git-ignored token file, then verify
with prowl_list_tools.

4 — Verify the connection

Ask your agent to run prowl_list_tools. It is free and returns all 448 tools grouped by category — if you see them, you are connected.

Server health needs no auth at all:

curl https://prowl.chat/mcp/health

For an end-to-end smoke test that costs nothing, call a $0 utility tool:

prowl_call_tool
  tool_name: extract_domain_from_url
  params: {"url": "https://www.stripe.com/pricing"}

In the web app, Step 3 “Verify connection” watches for your agent’s first tool call and has a Check now button.

5 — Run your first report, then export it

Call prowl_analyze with a query and an execution mode. Keep the same session_id across calls to preserve context.

• basic — roughly 20–100 tool calls, about 30–90 seconds. Available on the free tier.
• deep — roughly 40–300 calls, about 3–5 minutes. Requires a plan.
• max — roughly 60–400 calls, about 5–10 minutes. Requires a higher plan.

Asking for deep or max without an active subscription does not fail — the run is downgraded to basic and you get an explicit notice before any money moves.

Long runs can go asynchronous: prowl_start_session returns immediately, then poll prowl_session_status until it completes and read it with prowl_get_session.

To export, run prowl_generate_artifact (infographic, pdf or pptx) or prowl_export_report (markdown or html). Both require a prior prowl_analyze on the same session_id — they work from that session’s cached report.

Funding your wallet

The $5 starter credit covers light testing. To add more, open MCP Home → Wallet and pick a top-up pack ($10, $50 or $200) — checkout runs through Stripe and the credit lands on your extra balance. Subscriptions add a monthly credit and unlock the deep and max tiers.

Every run reserves a hold and then settles down to what it actually used: you are billed at most the hold, never more.