Agent-native scouting via MCP
Connect ScoutingAPI as a Claude/Cursor MCP connector and scout stays, compare prices and check availability conversationally — no keys pasted.
What it does
The job it gets done
Give an AI agent first-class access to accommodation data without pasting an API key into the model. Connect ScoutingAPI’s native MCP server once (OAuth 2.1 + PKCE), and the agent gains seven read-only tools — search stays, check availability, get a listing, get a price, compare prices across OTAs, get reviews, and poll async jobs — that it can call in the flow of a conversation over one unified schema. It is the agent-native path the scraping-template catalogs cannot match.
How it works
Step by step
- 1
Add ScoutingAPI as a custom MCP connector
In Claude.ai (or Claude Desktop / Cursor) add a custom connector with the server URL https://mcp.scoutingapi.com. No API key is pasted into the agent.
- 2
Authorize once (OAuth 2.1 + PKCE)
Complete the browser OAuth prompt — with dynamic client registration — to link the connector to your ScoutingAPI account and its credit balance. Usage is metered to that account.
- 3
Scout in the conversation
Ask the agent to find stays, compare a property across OTAs, or check availability. It calls search_stays, compare_prices, check_availability and the other read-only tools directly.
- 4
Get answers in the flow of the chat
The agent reads the unified-schema results and answers inline — the cheapest cross-OTA rate, bookable dates, or a ranked shortlist — no glue code, no keys in the prompt.
What you'll need
Two ways to run it
Using an AI agent
Just your ScoutingAPI key. Your agent is the brain — it calls the endpoint and reports the result itself, so there's no separate AI key to buy. Install the SKILL.md below.
Using n8n
A running n8n instance, the workflow JSON below, a ScoutingAPI Header Auth credential, and a delivery credential (e.g. Slack) — or swap the last node. The transform is plain JavaScript, so no LLM key is required.
Try it live
Run the core call now — sandbox, 0 credits
Exercise the endpoint this workflow is built on against the deterministic sandbox — a real 200, zero credits, no sign-up.
/v1/searchRuns against the deterministic Airbnb + Booking sandbox fixtures — a real 200, zero credits.
Run the call to see a live, deterministic sandbox response.
curl -sS "https://api.scoutingapi.com/v1/search" \
-H "Authorization: Bearer $SCOUTINGAPI_KEY"Code samples
The same call, every way that matters
// Claude Desktop / Cursor — claude_desktop_config.json (or add as a custom connector on Claude.ai)
{
"mcpServers": {
"scoutingapi": {
"url": "https://mcp.scoutingapi.com"
}
}
}
// You authorize once in the browser (OAuth 2.1 + PKCE) — no key is pasted into the agent.Get it for your platform
Download & import
Typical cost
from 5 credits per platform per run — scales with the results returned (v3 per-result).
search· from 5 credits per platformprice-compare· 22 credits per callavailability· from 5 credits per callprice· 3 credits per calllisting· 3 credits per callreviews· from 5 credits per call
Failed & empty calls are always free; sandbox calls cost 0. Live costs are on the pricing page.
FAQ
Common questions
Get your free key — no card
300 credits, the full API, the sandbox, and the native MCP server. Power this workflow and more.
No credit card · failed calls cost 0 credits