10 million investors are talking. Your AI can finally listen.

Here's how to actually connect your agent to the Stocktwits signal: two steps, then two worked setups you can lift line for line.

Two setups you can lift line for line, one you'd build over a weekend, one you'd run from your desk.

Example A: The weekend build

Claude Code · Stocktwits MCP · Alpaca MCP · paper mode

Give Claude Code a signal and a broker on any always-on computer, and let it paper-trade while you sleep. Alpaca's MCP is live today; the Stocktwits MCP connects the same way.

~/agent · zsh

# 1) Add the signal layer: Stocktwits MCP
claude mcp add stocktwits --transport http https://mcp.stocktwits.com/mcp

# 2) Add the execution layer: Alpaca MCP (live today, endpoint per Alpaca's MCP docs)
claude mcp add alpaca --transport http https://mcp.alpaca.markets \
  --env ALPACA_MODE=paper

# 3) Talk to your agent in plain English
claude "Every 15 minutes, check get_trending_symbols on Stocktwits.
For any name with EXTREMELY_HIGH message volume AND >80% bullish sentiment,
read the recent messages, then place a small PAPER order on Alpaca and
log your reasoning. Never touch real money. Notify me before anything live."

What happens next: your agent sits there reading the crowd all day, paper-trading the loud, high-conviction names, and pinging you with its reasoning, zero real dollars at risk.

Here's one real pass of that loop, run against the live MCP on July 22, 2026 — minutes after Alphabet's Q2 earnings hit the tape:

▸ live output · stocktwits-mcp · captured jul 22 · 16:41 ET

agent ❯ get_trending_symbols()

#1▼ 3.61%

$TSLATesla Inc

$365.25

1.1M watching26.8M volSTOCK

#2▼ 6.05%

$SPCXSpaceX

$116.07

90.6K watching90.7M volSTOCK

#3▼ 2.01%

$NOWServiceNow Inc

$100.01

42.4K watching39.5M volSTOCK

#4▼ 1.45%

$GOOGLAlphabet Inc Class A

$342.12

219.5K watching32.2M volSTOCK

#5▼ 1.07%

$GOOGAlphabet Inc Class C

$342.50

310.1K watching20.1M volSTOCK

#6▼ 7.44%

$QSQuantumScape Corp

$5.60

66.6K watching28.1M volSTOCK

#7▼ 0.24%

$IBMInternational Business Machines Corp.

$210.00

69.3K watching13.5M volSTOCK

#8▼ 11.30%

$PATHUiPath Inc

$10.68

33.1K watching105.6M volSTOCK

#9▲ 2.75%

$AVGOBroadcom Inc

$397.13

70.5K watching16.4M volSTOCK

#10▼ 3.25%

$TXNTexas Instruments Inc.

$281.82

13.1K watching10.4M volSTOCK

earnings night: the whole board is reporting names, and the watch name sits at #4. read it directly.

agent ❯ get_symbol_pulse("GOOGL")

$GOOGLNASDAQ

Alphabet Inc Class A

$342.29-1.40%

Watchers

219,506

Msg volume 1D

33.5K -16.13%

Message sentiment-11.74% vs 1W

72% Bullish28% Bearish

72% bullish, volume HIGH but not EXTREMELY_HIGH — neither bar met. no paper order this pass. reasoning logged.

Example B: The desk setup

Claude Desktop · Stocktwits MCP · Robinhood agent account

Prefer a GUI? Wire the same signal into Claude Desktop with a config snippet, then connect Robinhood's sandboxed agent account for execution. Robinhood opened to agents in May 2026.

claude_desktop_config.json

{
  "mcpServers": {
    "stocktwits": {
      "comment": "No client ID or secret — sign in with your Stocktwits account.",
      "url": "https://mcp.stocktwits.com/mcp"
    },
    "robinhood": {
      "comment": "Live today, connects your separate, pre-funded agent account. Endpoint per Robinhood's agent docs."            ,
      "url": "https://agent.robinhood.com/mcp"
    }
  }
}

Then, in Claude Desktop:

❯ "Read the Stocktwits crowd on my watchlist. If anything is trending with strong bullish conviction, show me the actual messages and a one-line take. If I say go, place it through my Robinhood agent account, and remind me I have to approve it."

What happens next: you get a desk-side analyst that reads 10 million investors for you, surfaces the loud names with the receipts, and never trades until you say the word.

Broker endpoints are illustrative — follow each provider's docs for current URLs and auth. The live outputs above are real MCP responses captured July 22, 2026, minutes after Alphabet's Q2 earnings; prices, sentiment, and posts move in real time.

Point your agent at the signal in an afternoon.

Example A: The weekend build

Example B: The desk setup