// vercel ai sdk

Market data tools for
the Vercel AI SDK.

Register TickerDB methods as AI SDK tools. Your language model gets access to latest EOD market summaries, watchlists, and ticker event history — grounding financial responses in real data.

try for free docs

// setup

Install the dependencies.

Add the TickerDB SDK alongside the Vercel AI SDK.

terminal

npm install tickerdb ai @ai-sdk/openai

Set TICKERDB_KEY in your environment variables and pass it into the client when you initialize it.

// usage

Register tools. Stream responses.

Define TickerDB methods as AI SDK tools with typed parameters. The model calls them automatically when it needs market data.

Tool Registration lib/market-tools.ts

import { TickerDB } from "tickerdb";
import { tool } from "ai";
import { z } from "zod";

const client = new TickerDB({ apiKey: process.env.TICKERDB_KEY! });

export const marketTools = {
  getSummary: tool({
    description: "Get a market summary for a ticker",
    inputSchema: z.object({
      ticker: z.string().describe("Stock ticker symbol"),
    }),
    execute: async ({ ticker }) => {
      const { data } = await client.summary(ticker);
      return data;
    },
  }),
  getSummaryEvents: tool({
    description: "Ticker event history via summary field/band filters",
    inputSchema: z.object({
      ticker: z.string().describe("Stock ticker symbol"),
      field: z.string().default("rsi_zone"),
      band: z.string().default("deep_oversold"),
    }),
    execute: async ({ ticker, field, band }) => {
      const { data } = await client.summary(ticker, { field, band });
      return data;
    },
  }),
};

API Route app/api/chat/route.ts

import { openai } from "@ai-sdk/openai";
import { streamText } from "ai";
import { marketTools } from "@/lib/market-tools";

export async function POST(req) {
  const { messages } = await req.json();

  const result = streamText({
    model: openai("gpt-4o"),
    messages,
    tools: marketTools,
  });

  return result.toDataStreamResponse();
}

// agent workflow

Multi-step analysis.

Your model can chain tool calls — get a full summary, use summary event mode for historical context, then pull another ticker for comparison. Each call returns market-state data the model understands without extra prompting.

getSummary(ticker: "INTC")

Full categorical snapshot for the ticker

getSummaryEvents(ticker: "INTC", field: "rsi_zone", band: "deep_oversold")

Historical ticker events with aftermath

getSummary(ticker: "AMD")

Compare by pulling a second summary

// state changes

Track state changes effortlessly.

Most market data APIs return point-in-time snapshots. TickerDB tracks state transitions — your agent sees what changed, not just what is.

json

{
  "timeframe": "daily",
  "run_date": "2026-03-28",
  "changes": {
    "AAPL": [
      {
        "field": "rsi_zone",
        "from": "neutral",
        "to": "oversold"
      },
      {
        "field": "trend_direction",
        "from": "uptrend",
        "to": "downtrend"
      }
    ]
  },
  "ticker_context": {
    "AAPL": {
      "last_changed_date": "2026-03-28"
    }
  },
  "tickers_checked": 2,
  "tickers_changed": 1
}

// what your agent sees

What your agent sees.

Every tool returns categorical facts — not raw OHLCV data. Your agent can branch on "oversold" without needing to know what RSI > 70 means.

json

{
  "ticker": "NVDA",
  "data_status": "eod",
  "as_of_date": "2026-04-11",
  "trend": {
    "direction": "strong_uptrend",
    "ma_alignment": "aligned_bullish"
  },
  "momentum": {
    "rsi_zone": "overbought",
    "macd_state": "expanding_positive",
    "direction": "accelerating"
  },
  "volatility": {
    "regime": "normal",
    "regime_trend": "stable"
  },
  "fundamentals": {
    "valuation_zone": "fair_value",
    "pe_vs_historical_zone": "premium",
    "last_earnings_surprise": "beat"
  }
}