Buzzberg MCP

Give your AI agent the market intelligence of a $100M fund.

Your agent can track 200+ high-signal finance voices in real time. Buzzberg extracts 100+ trade ideas every day across X, YouTube, Reddit, newsletters, and other public market sources, then links each thesis to its source.

2,800+Tracked market sources
280K+Extracted trade ideas
5,600+Speaker profiles
Works with Claude Cursor Codex OpenClaw

The missing layer of stock research.

Generic LLM stock research misses things: even with web search on, an agent skims headlines, and the nuance that makes or breaks a thesis slips through. Buzzberg closes that gap — for every ticker, your agent gets every mention from hundreds of tracked voices across X, YouTube, Reddit, and Substack. Many authors mean many angles: theses, risks, and disagreements side by side — if it moved the stock, one of these voices caught it. And beyond the narrative, your agent reads the tape around it: sentiment per ticker, mention counts and their momentum, how the tone on X and elsewhere is shifting — plus every earnings-call mention and whether top funds hold it in their latest 13F filings.

Plain LLM
Training data plus a quick web pass: fundamentals, headlines, and whatever ranked first — one generic answer.
LLM with Buzzberg MCP
Every tracked mention of the ticker across X, YouTube, Reddit, and Substack — theses, risks, and disagreements from hundreds of voices — plus sentiment, mention momentum, and tone shifts, each linked to its source.

What your agent can do with it.

A fresh brief every morning.

Each morning your agent hands you the newest trade ideas, the trending trades, and the narratives driving the market — before you open a single feed.

Scan all tracked voices from the last 24 hours; state the data window and sources up top. First, new trade ideas: long or short calls on tickers new to the tape or barely mentioned in recent months — entry price, price 7 and 30 days before the call, YTD, the thesis, and a measurable invalidation trigger. Then, trending trade ideas: tickers with the most independent long calls — who's behind them, the shared thesis, and price now vs 7 days, 30 days, and YTD. Finish with today's most-mentioned tickers. Skip watch-only mentions; flag conflicts of interest and crowded run-ups.

Deep dive any ticker.

Bull and bear theses, risks, disagreements, sentiment and attention — plus earnings-call mentions — distilled into one sourced brief with the poker math on top.

Deep dive NVDA. Open with what the business actually does, in plain words. From Buzzberg, read every tracked mention — full text from X and Reddit, Buzzberg research notes for YouTube, Substack, and podcasts, and earnings-call mentions where available: the strongest bull and bear theses with the authors' own numbers and risks, how the thesis evolved over 90 days, whether the voices are independent or one community echoing itself, and a coverage map — who talks about it and who is silent. From your other tools: price, market cap, EV, and five years of revenue, margins, and FCF minus SBC. Then the poker math: a reverse DCF — what growth today's price implies; bear, base, and bull with probabilities, expected value, and risk/reward against a 2–3:1 bar. Verdict: position, watchlist, or pass — plus three measurable thesis-kill triggers.

Spot crowding before the crowd.

Chart mindshare and sentiment over time: see whether a ticker is overcrowded or attention is only starting to build — and how the tone is shifting.

Chart NVDA attention over the last 180 days — render the charts, don't just describe them: weekly price, mention counts, and average sentiment from the Buzzberg timeseries, plus where NVDA sits on today's most-mentioned leaderboard. Verdict: overcrowded, or attention just starting to build? Judge it by mentions vs their own 180-day range, whether sentiment confirms or diverges from price, and who leads whom. End with what would mark the turn.

Backtest your own feed.

Pick the 10–20 voices you actually trust, load their trade ideas and mentions, and test the strategy you've been eyeballing — against real history.

Replay any speaker's or podcast's history.

The full timeline of every call: how views changed, when conviction flipped, who was early — and who just followed the crowd.

What does the All-In Podcast think about AI and NVDA, and how has the view evolved? Build a dated timeline from their tracked episodes: each shift in thesis with its dated note, hosts vs guests, the direction mix over time, and what they've been right or wrong about so far. End with the current stance in one line.

Show aleabitoreddit's full SIVE history: the first call, every flip, how conviction changed over time, and what happened to the price after each call. End with the current stance in one line.

Setup

Connect in two minutes.

ChatGPT and Claude connect directly through Buzzberg OAuth. No Terminal, package install, or copied API key is required.

03 · OAuth

Sign in, then ask

Select Create in ChatGPT or Connect in Claude. Buzzberg opens in your browser; sign in and approve access. Then start a new chat: choose Developer mode → Buzzberg in ChatGPT or enable Buzzberg from Connectors in Claude.

OAuth completes the connection without a personal API key. ChatGPT custom MCP apps currently run on the web; Claude's remote connector follows the same account across web, Desktop, and mobile.

GitHub docs
Claude Code, Codex, OpenClaw, and existing API-key setups

Claude Code

Add the Streamable HTTP server, then open /mcp in Claude Code and follow the browser sign-in.

claude mcp add --transport http buzzberg https://mcp.buzzberg.ai/mcp

Claude Code stores and refreshes the OAuth connection for you.

Codex

Already installed Buzzberg in ChatGPT? Open /apps in Codex to reuse that app authorization. Use direct MCP only when you intentionally want a separate local CLI or IDE connection.

/apps

# Separate local connection only
codex mcp add buzzberg --url https://mcp.buzzberg.ai/mcp
codex mcp login buzzberg

Existing legacy key

Use this path only if your older client setup already has a bzb_... key. New personal keys are no longer issued.

Sign in to manage legacy keys

Existing bzb_ keys, /sse connections, and installed helper configs continue to work and remain revocable in Profile.

Claude Desktop manual fallback

Existing local Claude Desktop setups may keep the pinned bridge. New users should use the OAuth connector above.

macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
Windows: %APPDATA%\Claude\claude_desktop_config.json
{
  "mcpServers": {
    "buzzberg": {
      "command": "npx",
      "args": [
        "-y",
        "[email protected]",
        "https://mcp.buzzberg.ai/mcp",
        "--transport",
        "http-only",
        "--header",
        "X-API-Key:bzb_YOUR_KEY_HERE"
      ]
    }
  }
}

Keep X-API-Key: exactly as shown, with no space after the colon. Do not share screenshots of this config. If the legacy key stops working, remove this fallback and reconnect through OAuth; replacement keys are not issued.

Legacy SSE with a key

Existing Claude Code and older-client configurations can keep using the SSE endpoint unchanged.

export BUZZBERG_MCP_API_KEY="bzb_YOUR_KEY_HERE"
claude mcp add --transport sse buzzberg https://mcp.buzzberg.ai/sse \
  --header "Authorization: Bearer $BUZZBERG_MCP_API_KEY"

OpenClaw

Use Streamable HTTP and store the key in the client's secret or environment-variable field.

URL    https://mcp.buzzberg.ai/mcp
Header Authorization: Bearer bzb_YOUR_KEY_HERE
Buzzberg MCP

Start with one market question.

Connect Buzzberg with OAuth, then start with a ticker, speaker, narrative, or market-wide question.

Connect your agent