For AI assistants

Put the record inside your AI assistant

Tapeline runs a Model Context Protocol server, so Claude, ChatGPT, Cursor and any other MCP-compatible assistant can read our scores and our published track record directly — rather than paraphrasing a marketing page. It’s free, needs no account and no API key, and it returns exactly what an anonymous visitor can already see on this site.

Connect it

Server URL
https://api.tapeline.io/mcp
  1. 1
    Claude (Desktop or Code)
    Settings → Connectors → Add custom connector, and paste the URL above. In Claude Code: claude mcp add --transport http tapeline https://api.tapeline.io/mcp
  2. 2
    ChatGPT
    Settings → Connectors → Add, then paste the URL. Available where custom connectors are enabled for your plan.
  3. 3
    Cursor, Windsurf, or anything config-driven
    Add the block below to your MCP config file.
{
  "mcpServers": {
    "tapeline": {
      "url": "https://api.tapeline.io/mcp"
    }
  }
}

What it can answer

get_ticker_scoreThe current six-factor score, signal, confidence and one-line reason for any covered US ticker.
get_daily_picksToday's published top 10 — the same rows an anonymous visitor sees on the site.
get_track_recordThe full published record: entries logged, sessions tracked, share that beat SPY, median alpha — with the sample-size qualifier attached.
get_ticker_recordEvery time a ticker was published in the top 10, and how each of those picks actually resolved. Losses included.
search_tickersResolve a company name to a covered symbol.

Try asking your assistant: “What does Tapeline score NVDA, and why?” or “How has Tapeline’s published record actually done against SPY?”

What it will tell you about us

The same thing the site does, including the unflattering part. Every daily top-10 pick is written to a public record the moment it prints and is never re-ranked, edited or removed. One session later each pick’s realised move is compared against SPY and appended — losses and all.

At the current sample the picks do not beat SPY, and the server says so: every response carrying a performance number also carries the qualifier that at this sample size the values do not distinguish the ranking from chance. The numbers are read live on every call, so an assistant quotes today’s record rather than a figure that was true last month.

Descriptive scoring only — not investment advice, price targets or forecasts.

See the public recordHow the score works