ai.cabrini/market-data
Provides US stock market data for AI agents, including intraday and daily bars, SEC fundamentals, filings, and insider data, with pay-per-query via USDC on Base.
README
cabrini
US stock market data for AI agents. 23 years of intraday and daily bars, SEC fundamentals, filings, and insider data — every US equity from 2003 to present.
Pay per query with USDC on Base (x402). No API keys, no subscriptions, no signup.
Install
pip install cabrini
Quick start
from cabrini import Cabrini
c = Cabrini(private_key="0x...") # any Base wallet with USDC
# Intraday bars (pct from daily open) — $0.025
bars = c.query("AAPL", "2024-01-15")
# Daily OHLCV + VWAP (absolute prices) — $0.001/year
daily = c.daily("TSLA", "2024-01-01", "2024-03-31")
# SEC fundamentals — $0.02
fins = c.fundamentals("NVDA")
# Full research brief — $0.25
brief = c.brief("MSFT")
LangChain
from cabrini import get_langchain_tools
from langchain_openai import ChatOpenAI
from langgraph.prebuilt import create_react_agent
tools = get_langchain_tools(private_key="0x...")
agent = create_react_agent(ChatOpenAI(model="gpt-4o"), tools)
result = agent.invoke({"messages": [
{"role": "user", "content": "What was NVDA's trading volume on the day of their last earnings?"}
]})
CrewAI
from cabrini import get_crewai_tools
from crewai import Agent, Task, Crew
tools = get_crewai_tools(private_key="0x...")
analyst = Agent(
role="Financial Analyst",
goal="Analyze stock performance using real market data",
tools=tools,
)
task = Task(
description="Compare AAPL and MSFT intraday volatility on 2024-06-15",
agent=analyst,
)
Crew(agents=[analyst], tasks=[task]).kickoff()
MCP (Claude, Cursor, etc.)
Point any MCP client at https://cabrini.ai/mcp:
{
"mcpServers": {
"cabrini": {
"url": "https://cabrini.ai/mcp"
}
}
}
All endpoints
| Method | Price | Description |
|---|---|---|
query(ticker, date) |
$0.025 | Full trading day of intraday bars |
daily(ticker, start, end) |
$0.001/year | Daily OHLCV + VWAP — the absolute prices |
batch(tickers, date) |
$0.02/ticker | Several tickers, one date, no limit |
range(ticker, start, end) |
$0.01/trading day | Multi-day intraday, no limit |
bars(ticker, date, interval) |
$0.015/day | Resampled intraday, 3-240 min |
scan(date, **criteria) |
$0.10 | Screen every US stock; needs >= 1 criterion |
tickers(date) |
$0.005 | List traded tickers |
company(ticker) |
$0.005 | Company profile from SEC EDGAR |
fundamentals(ticker) |
$0.02 | SEC quarterly data |
filings(ticker) |
$0.01 / $0.05 | SEC filing index; +extracted section text |
insiders(ticker) |
$0.02 | Insider transactions (Form 4) |
brief(ticker) |
$0.25 | Joined research brief |
Prices are quoted live in each 402 response and the client pays whatever the server
asks — this table is documentation, not the source of truth.
Output format
Intraday methods (query, range, batch, bars) return fractional change from the
daily open, not price levels:
{"window_start": "2024-01-02T14:30:00", "timestamp": 1704204600000000000,
"pct_open": 0.0, "pct_high": 0.0012, "pct_low": -0.0003, "pct_close": 0.0008,
"volume": 47000, "transactions": 312}
pct_x = (bar_x - day_open) / day_open, so 0.0012 is +0.12%.
daily() carries the absolute levels — open, high, low, close, volume, transactions and
VWAP. Combine the two to reconstruct prices:
day = c.daily("AAPL", "2024-01-02", "2024-01-02")["data"][0]
bars = c.query("AAPL", "2024-01-02")["data"]
close_price = day["open"] * (1 + bars[-1]["pct_close"])
Use daily() rather than a third-party open: our reference is the first bar of the
session and includes pre-market, so an external 09:30 open will not reconcile exactly.
How payment works
Every paid request uses x402 — an open protocol for HTTP micropayments:
- Client sends request → server returns
402with aPAYMENT-REQUIREDheader - Client signs a USDC transfer authorization (EIP-3009)
- Client replays request with
X-PAYMENTheader containing the signed authorization - Cloudflare edge worker verifies signature, submits to Base, forwards to origin
- Origin returns data
The Cabrini client handles all of this automatically. You just need a wallet with USDC on Base.
Get USDC on Base
- Bridge from Ethereum: bridge.base.org
- Buy directly: Coinbase → send USDC to your wallet on Base network
- Faucet (testnet): not needed, mainnet USDC is cheap ($0.025/query)
Links
- Homepage: https://cabrini.ai
- API docs: https://cabrini.ai/docs
- Agent guide: https://cabrini.ai/agents
- MCP endpoint: https://cabrini.ai/mcp
<!-- mcp-name: ai.cabrini/market-data -->
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。