Skip to content

Scout for developers and agents

Scout's data is open: a JSON API, an MCP server for AI assistants, and in-page tools for browser agents. All three are free and read-only, and they need no sign-up, API key or token. They return the same public company data the site shows, which refreshes once a day.

Quickstart

No authentication: call the API directly. The base URL is https://scout.emre.computer/v1 and the machine-readable description is the OpenAPI 3.1 document (also listed in the API catalog).

curl "https://scout.emre.computer/v1/companies?sector=ai&range=7d"

The API is read-only, so every call is safe to try against production; there is no separate sandbox to set up.

Endpoints

  • GET /v1/companies: list and search companies, 100 per page. Query parameters, all optional:
    • q: search names and websites (max 60 characters)
    • sector: one of ai, quantum, hardware, robotics, biotech_health, space_defense, climate_energy, crypto, fintech, dev_tools_infra, consumer, other
    • range: 24h, 3d, 7d or all (default), by when Scout found the company
    • sort: newest (default) or growth (X followers gained over 7 days)
    • cursor: the previous response's nextCursor
  • GET /v1/companies/{handle}: one company by X handle, or 404 company_not_found.
  • GET /v1/stats: companies found on each of the last 7 days, optionally for one sector.
  • GET /v1/sectors: the sector slugs and their names.

Example response (illustrative values):

curl "https://scout.emre.computer/v1/companies/acme"

{
  "handle": "acme",
  "name": "Acme",
  "sector": "ai",
  "sectorLabel": "AI",
  "bio": "…",
  "websiteUrl": "https://acme.example/",
  "followers": 1234,
  "followerGrowth7d": 56,
  "foundDaysAgo": 2,
  "funding": { "round": "Seed", "amountUsd": 3000000, "date": "2026-10-01", "totalRaisedUsd": 3000000 },
  "xUrl": "https://x.com/acme",
  "scoutUrl": "https://scout.emre.computer/?c=acme"
}

Fields are null when Scout does not know them. Every field is described in the OpenAPI document.

Pagination

Lists are cursor-based. A response has companies, total, hasMore and nextCursor, plus a Link: <…>; rel="next" header. Repeat the same request with cursor=<nextCursor> until nextCursor is null. Cursors are opaque; pass them back unchanged.

Errors

Errors are RFC 9457 application/problem+json objects. Each has a stable code to branch on, a detail, and a hint that says how to fix the request.

HTTP/1.1 400 Bad Request
Content-Type: application/problem+json

{ "type": "https://scout.emre.computer/developers#errors", "title": "`range` must be one of: 24h, 3d, 7d, all.", "status": 400,
  "code": "invalid_parameter", "detail": "`range` must be one of: 24h, 3d, 7d, all.", "hint": "Example: range=24h" }

Codes: invalid_parameter, unknown_parameter, invalid_cursor (400); company_not_found, not_found (404); rate_limited (429); upstream_unavailable (502, retry); internal_error (500).

Rate limits

About 120 requests per minute per client IP, for the API and the MCP server together (a best-effort limit to stop runaway loops, not a quota). Every response has IETF RateLimit and RateLimit-Policy headers (and X-RateLimit-* equivalents). Over the limit you get a 429 with Retry-After. The data changes once a day, so caching responses for a few minutes costs nothing.

Versioning and deprecation

The major version is in the path (/v1) and in the API-Version response header. New fields, endpoints and enum values can be added within v1, so ignore fields you don't know. Breaking changes ship as a new major version. A version being retired is announced at least 90 days ahead, with Deprecation and Sunset headers on its responses and a note on this page.

MCP server

Connect an MCP client (Claude, ChatGPT, Cursor, …) to https://scout.emre.computer/mcp. It uses the Streamable HTTP transport and needs no authentication. Tools: search_companies (it renders an interactive list in hosts that support MCP Apps), get_company, get_discovery_stats and list_sectors. All tools are read-only. See the server card.

claude mcp add --transport http scout https://scout.emre.computer/mcp

WebMCP

On the homepage, browsers with WebMCP expose the same actions to their built-in agent through document.modelContext: search companies, get a company, and show a filtered list or a company on the page.

Also for agents: llms.txt, the agent skill, and the ARD catalog. Questions or problems: see contact.