MCP server · Claude, Cursor, Claude Code

Bring sources into your market research.

Search market topics, connect company context, and review how a story developed. Configure AlgoThesis in each compatible MCP client you use. Available data includes sources and reporting dates; missing or unlicensed fields stay unavailable.

2 anonymous research requests · then add an account API key · plan-based daily limits · MIT on npm

Install

01

Claude Desktop

Edit ~/Library/Application Support/Claude/claude_desktop_config.json (Mac) or %APPDATA%\Claude\claude_desktop_config.json (Windows). Add this block, then quit + relaunch:

{
  "mcpServers": {
    "algothesis": {
      "command": "npx",
      "args": ["-y", "algothesis-mcp"]
    }
  }
}
json
02

Claude Code (CLI)

One command:

claude mcp add --transport stdio algothesis -- npx -y algothesis-mcp
bash
03

Cursor / other MCP clients

In Cursor: Settings → MCP → Add Server → paste the same JSON block from step 1.

Use a client that supports local stdio MCP servers. Configuration steps vary by client.

04

Keep using it after the trial

Sign in, create a bot key in Settings, then add it to the MCP server's environment. The key is shown once; don't paste it into prompts or commit it to a repository.

{
  "mcpServers": {
    "algothesis": {
      "command": "npx",
      "args": ["-y", "algothesis-mcp"],
      "env": {
        "ALGOTHESIS_API_KEY": "algo_your_key_here"
      }
    }
  }
}
json

What you get

research_topicResearch any topic

Live Google Search + ticker mapping. Ask about people, events, macro themes — get a grounded summary with source URLs and matched company context.

Example: "What changed in the latest Federal Reserve announcement? Link to the source."

get_short_interestReal short-interest data

Reported short interest via Yahoo Finance for up to 20 tickers, when licensed data is available. Check each report date; these delayed figures do not predict a squeeze.

Example: "Compare reported short interest for NVDA and AMD, including report dates and any unavailable fields."

find_catalystsCheck upcoming catalysts

Earnings, filings, Fed meetings, and other scheduled events — by US ticker or for the overall market calendar.

Example: "What are the next catalysts for NVDA?"

company_recordWhat the stock did after the news — the record

Base rates from our own receipts: how many stories named this ticker, the median move at the first close after them, how often it went up, split by topic. Price fields appear only when licensed prior-trading-day closes are available. Historical association does not establish causation.

Example: "What is the record on NVDA — do stories on it actually move the stock?"

story_recordOne story, its receipt and its arc

Pass the UUID at the end of an AlgoThesis story link to review its available price receipt and accepted reports in the same developing story.

Example: "Open this AlgoThesis story and tell me what happened after it."

what_movedA trading day’s news and subsequent moves

Stories sorted by the first closing-price reaction of the company each named, when licensed receipts are available. This describes sequence, not the cause of a move.

Example: "What moved yesterday, and which story came first?"

Research prompts to try

Paste any of these into Claude after install. The model picks the tool.

What changed in the latest Federal Reserve announcement? Include source links.
Compare reported short interest for NVDA and AMD, including report dates.
Which companies are named in recent AI infrastructure coverage?
What are the next scheduled catalysts for NVDA? Verify their dates.
Summarize the available story record for NVDA and its limitations.
What moved yesterday, and which reports preceded the close?

Why this exists

A useful research answer should make its evidence inspectable. These tools connect your client to AlgoThesis research and public story records. Follow the source links and check the data dates. Provider outages and licensing limits can make a tool or field unavailable.

Source on npm · MIT · contributions welcome.