The US medtech market, machine-readable

Files

• index.json: machine-readable manifest of everything here
• signals-latest.json: buying signals from the last 14 days
• llms.txt: this site, for machine readers
• stats.json: aggregate history and archive scale (aggregates only)
• methodology.md: how the data is collected, dated, and windowed, what it holds about people, and what these files exclude
• signals-archive.json: the complete archive as one file, priced by the x402 catalogue

The files above are a rolling 14-day window, except the archive file, which is everything. The archive is also queryable free, a search or a company at a time, below.

Query it by MCP

The same data answers over the Model Context Protocol (Streamable HTTP). Connecting shows a sign-in page. Choose Continue with free access. HTTP routes below need no key.

https://data.whimbrelresearch.com

Free tools: latest_signals, signal_of_the_day, weekly_pulse, who_got_funded, find_signals, micro_brief, company_timeline, about_the_data, get_access. A first call after connect is signal_of_the_day or who_got_funded. weekly_pulse and latest_signals cover the recent window, including ClinicalTrials.gov device trial registrations, Breakthrough marketing authorizations, and HDE approvals. find_signals searches the whole archive, including the full trial-registration history, Breakthrough marketing authorizations, and HDE approvals. company_timeline does the same for one company. who_got_funded returns NIH awards. After that first call, ask weekly_pulse or latest_signals. If a Breakthrough manufacturer shows up, several events can name the same manufacturer. Ask for one full company brief on that company, not once per event. To subscribe, call get_access from a connected client. micro_brief and company_timeline are capped at 100 company lookups per caller per day. The server ships seven ready-made whimbrel_-prefixed prompts, so clients that surface prompts as slash commands get a command palette on connect.

Setup steps for Claude, Claude Code, Codex, Cursor, VS Code and other MCP clients: https://whimbrelresearch.com/connect/. Any client that speaks Streamable HTTP and OAuth sign-in can use the URL above directly; clients without OAuth read the HTTP routes below.

Query it by HTTP

The same handlers answer GET at the same origin, for readers that do not speak MCP:

/v1/subscribe?plan=<name> redirects to that plan's Stripe checkout, and /v1/key/{session} hands over the resulting API key once. With no plan named it sends a human to the setup page instead, because subscribing from inside a connected client hands the key to the app and never displays one.

Discovery and terms are free and stay free: the x402 catalogue lists every route, its price, and whether payment is currently switched on, and /v1/about states the terms in prose. Read the price from the catalogue rather than from this page.

The paid layer

A Whimbrel key unlocks the rest of the same server: deep_record, research_status, crm_export, billing_portal. A full company brief runs the research pipeline on any company you name and comes back with current leadership, contact routes, funding and milestones, every line quoted from the source it came from. research_status tracks the run and the month's remaining allowance; crm_export shapes a finished record into ready-to-import HubSpot or Salesforce files. The plans differ in one thing only, how many companies you research in a month, and the archive above stays free on all of them. get_access is free and needs no key: it returns a checkout link per plan, and the page after paying shows an API key once. Send the key as an Authorization: Bearer header. billing_portal handles card changes, plan changes and cancellation. What the paid layer is and who it serves: https://whimbrelresearch.com/connect/.