Docs

MCP Server

Ask AlphaScout from your AI assistant. The server speaks the Model Context Protocol, so a model can search companies, read their evidence and find comparables on its own.

Connect a Client

The server is at https://tryalphascout.com/api/v1/mcp. It uses the streamable HTTP transport and signs in with the same API key as the REST API, sent as a Bearer header. Create a key in Account → API, then add the server to your client:

Claude Code

claude mcp add --transport http alphascout https://tryalphascout.com/api/v1/mcp \
  --header "Authorization: Bearer as_live_…"

Cursor: save as ~/.cursor/mcp.json (or the project's .cursor/mcp.json). VS Code: .vscode/mcp.json; it asks for the key once and keeps it out of the file. Claude Desktop connects to remote servers with a header through the mcp-remote bridge, as shown.

Try It

Then ask in plain language, for example:

"Which fintech companies in Southeast Asia launched this week, and what is the evidence?"
"Find companies similar to company 1234 and explain the match."
"Why did this company score 58? Show me each rule."

You can also call the server directly. This lists three hidden gems in Europe:

curl

curl -s https://tryalphascout.com/api/v1/mcp \
  -H "Authorization: Bearer as_live_…" \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"list_hidden_gems","arguments":{"region":"europe","limit":3}}}'

Tools

Every tool is read-only and has the same filters as its REST counterpart. Lists default to 10 results (up to 50) so answers fit a model's context; ask for more with limit and offset.

Tool	What it does
list_hidden_gems	Ranked hidden gems: new launches nobody has backed yet, best first. Use this first when the user wants promising early companies. Every item has a gem score (0-100), labels, and the reasons behind the score with links to the evidence. Only new, un-backed launches qualify. Filter by sector, country or region, and use `since` to get only companies first seen after a date.
search_companies	Every tracked early company, filtered and sorted. Broader than hidden gems: all early companies AlphaScout tracks, including those without a gem score. Use it to look a company up by name or website (`q`), or to list companies that have evidence from several independent signal families (`min_families`).
get_company	One company's full profile with its evidence and funding. Returns the profile, every signal (each with its source link), identifiers and funding history for one company. Get the id from search_companies or list_hidden_gems.
similar_companies	Comparable companies, with the reasons each one matches. Finds companies that resemble one you already have, and says why (sector, business model, customers, market or a product that reads alike).
latest_releases	What companies just shipped, one entry per company. New products, directory listings, AI models, MCP servers and open-source repositories from the last days. Use `kind` to narrow it.
latest_signals	Raw evidence, newest first. The raw feed behind the rankings: launches, filings, hires, accelerator batches, grants and press, each linked to the company it belongs to and to where it was published.
scoring_rules	The scoring rules exactly as the engine applies them. The published weights per signal type and the gem adjustments. Use it to explain why a company scored as it did.

Protocol Notes

Transport: streamable HTTP, stateless. POST one JSON-RPC 2.0 message per request and read one JSON reply. There are no sessions and no server-sent event streams, so GET returns 405.
Protocol versions: 2025-06-18, 2025-03-26 and 2024-11-05. Methods: initialize, ping, tools/list, tools/call. No resources or prompts.
Batches aren't accepted (one message per request), so each tool call counts once against your limits.
Limits are the API's: 60 requests a minute and 10,000 a day per key. A 429 carries Retry-After.
A bad argument returns a tool result with isError: true and a message the model can act on, not a protocol error.

Security

The server only reads. Nothing a model asks can change your pipeline, shortlist, keys or billing.
Requests from a browser origin other than AlphaScout's are refused, and the server sends no CORS headers.
Treat a key in an assistant's config like any password; revoke it from the API page if a laptop is lost.
Company data is third-party content gathered from public sources. If you let an assistant act on it, treat it as untrusted input.