Vibe coding

Give an AI coding assistant the llms.txt file before it writes code against the client. Describe the analysis you want, then run the script it produces. If you want the assistant to fetch data and answer directly instead of writing code, connect it to the MCP server.

Choose an approach

• Context file: paste llms.txt into the conversation. The assistant writes Python that calls the xfinlink client, and you run the code. Use this for scripts, notebooks and repeatable analysis.
• MCP server: the assistant calls the API itself through 11 tools and answers from the JSON it gets back. Use this for questions you want answered in the chat. See MCP server.

Both use your API key and your plan's limits.

The llms.txt files

llms.txt holds the Quickstart and the Python client reference: every function with its arguments, return shape and errors, input formats, and runnable samples. Eight companion files each hold the reference pages for one group of HTTP endpoints. All are plain text (text/plain; charset=utf-8) and need no key.

• https://xfinlink.com/llms.txt: the Python client.
• https://xfinlink.com/llms-prices.txt: GET /v1/prices/{ticker}.
• https://xfinlink.com/llms-fundamentals.txt: GET /v1/fundamentals/{ticker}.
• https://xfinlink.com/llms-metrics.txt: GET /v1/metrics/{ticker}.
• https://xfinlink.com/llms-resolve.txt: GET /v1/resolve/{ticker}.
• https://xfinlink.com/llms-search.txt: GET /v1/search.
• https://xfinlink.com/llms-index.txt: index constituents and index events.
• https://xfinlink.com/llms-insiders.txt: GET /v1/insiders/{ticker}.
• https://xfinlink.com/llms-holdings.txt: holdings, manager search and manager holdings.

https://api.xfinlink.com/llms.txt serves the same file. Responses may be cached for up to one hour (Cache-Control: public, max-age=3600). See llms.txt for what each file contains.

$curl -s https://xfinlink.com/llms.txt | head -3

# xfinlink

Python client for US equity data: prices, financial statements, computed metrics, entity resolution, search, index constituents and membership changes, insider transactions, and institutional (Form 13F) holdings. ...

Load it with llm_context()

xfl.llm_context() fetches llms.txt, prints it, and returns it as a string. When the pyperclip package is installed, it also copies the text to your clipboard, ready to paste. It sends no API key.

import xfinlink as xfl

text = xfl.llm_context()
print(type(text).__name__)

# xfinlink

Python client for US equity data: prices, financial statements, computed metrics, entity resolution, search, index constituents and membership changes, insider transactions, and institutional (Form 13F) holdings. ...
...
https://xfinlink.com/llms-holdings.txt

str

The returned string is the same file that https://xfinlink.com/llms.txt serves.

Write the prompt

Put four things in the prompt: the reference, how the script gets the key, what your plan allows, and the task. A template:

You are writing Python with the xfinlink client (pip install xfinlink, version 0.13.1).
The client reference is below. Start the script with xfl.set_api_key("YOUR_API_KEY").
My plan is Free: one ticker per call, and the last 12 months of prices, fundamentals and metrics.
Fundamentals money values are in millions of the currency in each row's `currency` field (USD on a `USD` row). Prices are USD per share.

[paste the output of xfl.llm_context() here]

Task: ...

On a paid plan, state that plan instead, for example My plan is Pro: up to 100 tickers per call and full history. See Plans and limits.

If your assistant can open web pages, give it the URL https://xfinlink.com/llms.txt instead of pasting the text. For work against one HTTP endpoint without the Python client, paste that endpoint's companion file.

Example prompts

Each example shows the task you type after the template, the code an assistant should produce, and the real output of that code. The examples run on every plan.

Compare revenue across companies

Task: compare quarterly revenue for Apple, Microsoft and Alphabet for the quarters ending since October 2025, in billions of USD.

1
2
3
4
5
6
7
8
9
10
11import pandas as pd
import xfinlink as xfl

xfl.set_api_key("YOUR_API_KEY")

frames = [xfl.fundamentals(t, period_type="quarterly", start="2025-10-01", fields=["revenue"])
          for t in ["AAPL", "MSFT", "GOOGL"]]
df = pd.concat(frames)
df["revenue_bn"] = df["revenue"] / 1000
cols = ["ticker", "fiscal_year", "fiscal_period", "period_end", "revenue_bn"]
print(df[cols].round(1).to_string(index=False))

ticker  fiscal_year fiscal_period period_end  revenue_bn
  AAPL         2026            Q1 2025-12-27       143.8
  AAPL         2026            Q2 2026-03-28       111.2
  AAPL         2026            Q3 2026-06-27       109.4
  MSFT         2026            Q2 2025-12-31        81.3
  MSFT         2026            Q3 2026-03-31        82.9
  MSFT         2026            Q4 2026-06-30        90.0
 GOOGL         2025            Q4 2025-12-31       113.8
 GOOGL         2026            Q1 2026-03-31       109.9
 GOOGL         2026            Q2 2026-06-30       119.8

revenue arrives in millions of USD, so the code divides by 1,000. start filters on period_end, and each company's quarters end on its own fiscal calendar. The code makes one call per ticker, which works on every plan; on Pro and higher, one call can carry all three tickers.

Total return versus price return

Task: compute Apple's total return from the last close of 2025 to 25 September 2026, including dividends, and its price-only return, as percentages.

1
2
3
4
5
6
7
8
9
10import xfinlink as xfl

xfl.set_api_key("YOUR_API_KEY")

df = xfl.prices("AAPL", start="2025-12-31", end="2026-09-25")
daily = df["return_daily"].iloc[1:]           # drop the 2025-12-31 base row
total_return = (1 + daily).prod() - 1         # includes cash dividends
price_return = df["adj_close"].iloc[-1] / df["adj_close"].iloc[0] - 1
print(f"Total return: {total_return:.2%}")
print(f"Price return: {price_return:.2%}")

Total return: 25.80%
Price return: 25.46%

return_daily is the total daily return as a decimal, including cash distributions. adj_close is adjusted for splits, stock dividends and spin-offs but not cash dividends, so the gap between the two figures is the dividend contribution. The window starts on 2025-12-31 so that the first 2026 return has a base price. See Price fields.

A recycled ticker

Task: the ticker GM has belonged to more than one company. List every company that used it and the dates each one held the ticker.

1
2
3
4
5
6
7import xfinlink as xfl

xfl.set_api_key("YOUR_API_KEY")

info = xfl.resolve("GM")
for e in info["data"]["GM"]["entities"]:
    print(e["entity_id"], e["name"], e["ticker_valid_from"], e["ticker_valid_to"])

10001879 GENERAL MOTORS CO 2010-11-18 None
10001629 GENERAL MOTORS CORP 1962-07-02 2009-06-01

A ticker alone resolves to one company for your date window. resolve() lists every company that used the ticker, the current holder first, and an entity_id from that list reaches one company directly: xfl.prices(entity_id=10001629, period="max") returns General Motors Corporation's series, which ends on 2009-06-01 and needs a paid plan's full history. See Tickers and entities.

Check the generated code

Assistants reuse patterns from other libraries. Check these points before you trust a script:

• period and start or end cannot be combined; the client raises XfinlinkError.
• resolve() takes the date as date_; date= raises TypeError.
• Fundamentals money values are millions of each row's currency, and metrics money values millions of USD; prices market_cap and holdings value_usd are whole USD; ratios are decimals.
• insiders() sends include_amendments=false and holdings() sends include_amendments=true unless the code passes a value.
• insiders(), holdings() and manager_holdings() return at most 5,000 rows per call.
• Without start or period, prices(), fundamentals(), insiders(), holdings(), manager_holdings(), and metrics() with annual or quarterly, ask for the last 365 days; period="max" asks for full history. On Free, prices, fundamentals and metrics cover the last 12 months whatever start asks for.

Each point is shown with real output on the Python client reference.

Use the MCP server instead

Coding assistants that support MCP can call the API directly. In Claude Code:

$claude mcp add --transport http --scope user xfinlink "https://api.xfinlink.com/mcp?api_key=YOUR_API_KEY"

Setup for Cursor, VS Code, Claude, ChatGPT and others is on MCP server.