langchain-alphai

LangChain tools and a retriever for AlphAI — AI-scored, ticker-linked financial news and SEC Form 4 insider events, built for AI agents and trading bots.

Every article on the AlphAI feed is enriched before you see it: per-ticker impact analysis, one of 14 categories, and a 1-10 market-relevance score. The components here fetch and filter that feed — no scraping, no scoring of your own.

• AlphaAINewsSearch — the scored news feed with ticker / category / relevance filters
• AlphaAIInsiderNews — SEC Form 4 insider events with a structured who-sold-what block
• AlphaAITickerSentiment — 7-day bullish/neutral/bearish rollup for one ticker
• AlphaAIInsiderSummary — 30-day insider buy/sell rollup for one ticker
• AlphaAINewsRetriever — the feed as LangChain Documents for RAG pipelines

Install

pip install langchain-alphai

Requires Python 3.10+.

Authentication

Create an API key at alphai.io/developers — the free tier works without a card. Export it as ALPHAI_API_KEY, or pass api_key=... to any component.

export ALPHAI_API_KEY="ak_live_..."

Use the tools in an agent

from langchain.agents import create_agent

from langchain_alphai import (
    AlphaAIInsiderNews,
    AlphaAIInsiderSummary,
    AlphaAINewsSearch,
    AlphaAITickerSentiment,
)

agent = create_agent(
    model="claude-sonnet-4-5",
    tools=[
        AlphaAINewsSearch(),
        AlphaAIInsiderNews(),
        AlphaAITickerSentiment(),
        AlphaAIInsiderSummary(),
    ],
)

agent.invoke({"messages": [("user", "Are NVDA insiders selling, and does the news explain why?")]})

Each tool also works standalone:

from langchain_alphai import AlphaAINewsSearch

tool = AlphaAINewsSearch(min_relevance=7)  # instance defaults, overridable per call
tool.invoke({"symbol": "NVDA", "max_results": 5})

{
    "results": [
        {
            "uid": "788e477c66f3849b",
            "url": "https://...",
            "title": "Nvidia beats on data-center revenue",
            "summary": "Q2 revenue came in above consensus...",
            "source": "Example Wire",
            "source_domain": "example.com",
            "published_at": "2026-08-14T12:30:00+00:00",
            "tickers": ["NVDA"],
            "category": "earnings",
            "relevance_score": 8,
        }
    ]
}

Insider events add a structured insider block — side, shares, average price, total dollar value, who traded, and whether the sale was a pre-planned 10b5-1:

from langchain_alphai import AlphaAIInsiderNews

AlphaAIInsiderNews().invoke({"symbol": "NVDA", "min_relevance": 7})

On the insider feed the relevance score is deterministic from the event's summed dollar value, so min_relevance works as an "only large trades" dial.

Use the retriever for RAG

The query is a ticker symbol (crypto uses BTC-USD, foreign listings the Yahoo suffix, e.g. VOD.L); an empty query returns the market-wide feed.

from langchain_alphai import AlphaAINewsRetriever

retriever = AlphaAINewsRetriever(min_relevance=7, k=5)
docs = retriever.invoke("NVDA")

docs[0].page_content  # title + summary
docs[0].metadata  # tickers, category, relevance_score, url, source, ...

Filters

Async

Every tool implements ainvoke, and the retriever _aget_relevant_documents, over the SDK's native async client:

await AlphaAINewsSearch().ainvoke({"symbol": "NVDA"})

Rate limits

Limits are per AlphAI account, two-layer (per-minute burst + per-day volume): Free 20/min · 100/day, Basic 60/min · 10,000/day, Pro 150/min · 100,000/day. News-archive depth is tiered too (30/90 days / full archive). The underlying alphai-sdk retries 429s with backoff automatically.

Links

• Developer guide: https://alphai.io/developers
• API reference: https://api.alphai.io/api/schema/
• MCP server (same feed, for MCP-speaking agents): https://alphai.io/mcp

AlphAI output is AI-generated financial information for research, not investment advice — see alphai.io/terms.