DataXAPI MCP server: market data in your AI, connected with one prompt

Copy one prompt with your key in it, paste it into your agent, and ask. It reads one page and sets itself up: 5 tools, 207 endpoints on 21 platforms, from $0.0003 a call, and a failed call is never charged.

Sign up for the free trial
MCP addresshttps://api.dataxapi.com/mcp
  • A remote address: nothing to deploy, no platform account to sign in to
  • Give the key a daily and a monthly spending limit; your AI cannot go past it
  • Every call is in the console's call log, its source marked MCP
Setup steps for Claude Code, Cursor, Codex, n8n, Cherry Studio
Your AIconnected to DataXAPI
Which TikTok videos about camping coffee got the most likes this week?
find{ query: "search tiktok videos by keyword" }FreeFound tiktok.videos.search · $0.0020 a call
describe{ ids: ["tiktok.videos.search"] }FreeTakes query, sort, published_within · returns 30 posts a page
call{ capability: "tiktok.videos.search", query: "camping coffee", sort: "popular", published_within: "week", max_price: "0.01" }30 posts · compact view

Up to 30 post objects came back in the standard shape. Your AI answers from it and says what the call cost.

Each post comes back with
  • url
  • type
  • text
  • lang
  • published_at
  • author.name
  • metrics.likes
  • metrics.comments
  • metrics.shares
A call recorded on 2026-10-12 returned 30.
3 tool calls · 1 charged$0.0020
To connect
1 prompt$0. Finding and reading endpoints is free too
Tools
5 toolsonly one of them is charged
Behind them
207 endpoints21 platforms, one response shape
Answer size
≈ 20 KBcut on an item boundary; the rest is read free

Connected in three steps

No code to write

Sign up and get a key

Sign in with a 6-digit code sent to your email, or with Google. Free trial on sign-up: up to 333 calls at $0.0003. No card.

dx_…shown in full once, when it is made
Sign up free

Send this to your AI

It reads the setup page and connects the MCP server itself. Put your own key where it says dx_….

Read https://dataxapi.com/agents.md and set up DataXAPI in this agent. My API key is dx_…; use it for every DataXAPI call.

Just ask

Your AI finds the endpoint with find, runs it with call and answers with links. Every call is in the console's call log.

“What is trending on Douyin, Weibo and B站 right now?”
More questions

Prefer to do it by hand?

Every client connects to the same address, the key in a header. Give that key a daily spending limit, so a few extra calls cannot run past it.

Endpoint
https://api.dataxapi.com/mcp
Header
Authorization: Bearer dx_…
Transport
Streamable HTTP

Or one command, which writes the entry for Claude Code, Cursor, Codex and installs the base skill: dataxapi setup

What an agent reads to set itself up →
Claude Code1 command · ~/.claude.json
1

Put your key in the environment

Claude Code reads it from there. Add the line to your shell profile to keep it.

export DATAXAPI_API_KEY="dx_…"
2

Add the server

One command, or the same entry written into ~/.claude.json.

dataxapi setup --client claude-code
{
  "mcpServers": {
    "dataxapi": {
      "type": "http",
      "url": "https://api.dataxapi.com/mcp",
      "headers": {
        "Authorization": "Bearer ${DATAXAPI_API_KEY}"
      }
    }
  }
}
3

Restart Claude Code

Then ask it to list the DataXAPI tools. You should get the five below.

5 tools · only call is chargedSetup steps for Claude Code →

Cursor1 command · ~/.cursor/mcp.json
1

Put your key in the environment

Cursor reads it from there. Add the line to your shell profile to keep it.

export DATAXAPI_API_KEY="dx_…"
2

Add the server

One command, or the same entry written into ~/.cursor/mcp.json.

dataxapi setup --client cursor
{
  "mcpServers": {
    "dataxapi": {
      "url": "https://api.dataxapi.com/mcp",
      "headers": {
        "Authorization": "Bearer ${env:DATAXAPI_API_KEY}"
      }
    }
  }
}
3

Restart Cursor

Then ask it to list the DataXAPI tools. You should get the five below.

5 tools · only call is chargedSetup steps for Cursor →

Codex1 command · ~/.codex/config.toml
1

Put your key in the environment

Codex reads it from there. Add the line to your shell profile to keep it.

export DATAXAPI_API_KEY="dx_…"
2

Add the server

One command, or the same entry written into ~/.codex/config.toml.

dataxapi setup --client codex
[mcp_servers.dataxapi]
url = "https://api.dataxapi.com/mcp"
bearer_token_env_var = "DATAXAPI_API_KEY"
3

Restart Codex

Then ask it to list the DataXAPI tools. You should get the five below.

5 tools · only call is chargedSetup steps for Codex →

n8nThe MCP Client Tool node, attached as a tool to an AI Agent node of a workflow
1

Open the place MCP servers are added

The MCP Client Tool node, attached as a tool to an AI Agent node of a workflow

2

Fill in the fields

Each as written here. The key is typed into the client and kept by it.

FieldValue
Endpointhttps://api.dataxapi.com/mcp
Server TransportHTTP Streamable
AuthenticationBearer Auth
Credential for Bearer Auth → Bearer Tokenyour API key
3

Save, then ask it to list the tools

You should get the five below.

5 tools · only call is chargedSetup steps for n8n →

Cherry StudioSettings → MCP Servers → Add Server
1

Open the place MCP servers are added

Settings → MCP Servers → Add Server

2

Fill in the fields

Each as written here. The key is typed into the client and kept by it.

FieldValue
Namedataxapi
TypeStreamable HTTP
URLhttps://api.dataxapi.com/mcp
HeadersOne header: Authorization, with the value Bearer, a space, and your API key
3

Save, then ask it to list the tools

You should get the five below.

5 tools · only call is chargedSetup steps for Cherry Studio →

Any MCP client · One address and one header. A client that can read the environment takes the key from there; nothing has to be written into a file.

MCP
# Endpoint
https://api.dataxapi.com/mcp

# Header
Authorization: Bearer <dx_…>

What you can ask

Under each question: the endpoint your AI calls, and what one call costs

What is on Douyin's hot search board right now?

Find Xiaohongshu (RedNote) notes that review sunscreen.

Find TikTok videos about camping coffee.

Get the transcript of this YouTube video and sum it up.

What is trending on X (Twitter) in Japan today?

What do people on Reddit say about standing desks?

Use one of these, or your own words.
  • What is trending on Douyin, Weibo and B站 right now?
  • What did @trailbrew post on TikTok this week, and which video took off?
  • What are people praising and complaining about under this video?
  • Which coffee creators on Douyin and TikTok are worth a brief?
  • What ads are running for “running shoes” on Meta and TikTok?

One platform at a time: its questions, its endpoints and its skill package

What a session looks like

You ask in plain words. The AI does the rest by itself, and tells you what it spent.

It searches before it calls.

No endpoint names to remember.

find · Free

It sees the price first.

And can set the most it will pay for the call.

describe · Free

It gets an answer that fits.

Compact by default and cut at about 20 KB; read_more reads the rest at no charge.

call · $0.0020

It shows the bill.

Three tool calls, one charged.

3 tool calls · 1 charged · $0.0020

You set the limits

An AI that can spend needs a ceiling. There are three, and the server enforces them.

A spend limit on every key

Give your AI its own key with a daily, monthly or total budget. When it is used up, calls stop. Nothing else on your account is touched.

A ceiling on every call

With max_price, a call that could cost more is refused before anything is charged. A call that is sent twice with the same idempotency_key is charged once.

max_price: "0.01" · nothing if the call fails

Read-only, public data

DataXAPI reads what anyone can see on the platform. It cannot post, message, sign in to an account or change anything.

4 of 5 tools marked readOnly

The 5 tools your AI gets

findFinds the endpoint for a question.Up to 5 matches, each with its price and a ready-made call.no keyFree
describeReads up to 5 endpoints in full: parameters, fields, price.Also what an endpoint cannot return, and the most one call can cost.no keyFree
callRuns one endpoint and returns one page of data.The answer carries the charge, your balance and how to get the next page.needs a keyfrom $0.0003 · at the endpoint's own price · nothing if the call fails
read_moreRe-reads a result for 24 hours.The rest of a long answer, or the same result in another view.needs a keyFree
accountBalance, usage and receipts.Also this key's budget, and which endpoints are up (that part needs no key).needs a keyFree

Your AI can search and read the whole catalog before you have an account. Only call is charged, and only when data comes back.

Finding, describing, reading on, accountFree
A callfrom $0.0003
Errors and timeoutsNot charged

Charged per call from your balance. No subscription, and the balance does not expire. Top up from $10, by card or Alipay.

Clients

Claude CodeCursorCodexn8nCherry Studio
ChatGPTconnector
Claudeconnector

Any other client that speaks remote MCP connects with the address and a key.

Skills →

Questions

Before you connect.

What does it cost?

Connecting is free. Searching the catalog, reading an endpoint, re-reading a result and checking your account are free. A call costs the endpoint's listed price, from $0.0003, and nothing if it fails.

Do I need to install anything?

No. It is a remote server at one address. dataxapi setup only writes the entry into your client's configuration (Claude Code, Cursor, Codex) and installs the base skill; the prompt lets your agent do that for you.

Can my AI spend money without asking me?

call is not marked read-only, so a client that asks before running such tools will ask. Either way, the key's budget is the hard stop.

Will new platforms need a new server?

No. The 5 tools stay the same; a new endpoint shows up in find the day it ships.

What if an answer is too long for the context?

An answer is cut at about 20 KB, on an item boundary, and says how to read on. read_more reads the rest for 24 hours at no charge.

Does it work in ChatGPT, Claude?

Yes, on the web and in the desktop apps: add a custom connector with the address https://api.dataxapi.com/mcp. No key is typed in: you are asked to sign in to DataXAPI and approve, and there you choose the key its calls are charged through and that key's spending limit. ChatGPT: Settings → Apps & Connectors → turn on developer mode under Advanced → Create a connector: the MCP address, with OAuth as its authentication. Claude: On the web or in the desktop app: Settings → Connectors → Add custom connector, with the MCP address.

Connect once. Ask about any market.

Free trial on sign-up: up to 333 calls at $0.0003. Sign-up takes an email and no card. A failed call is never charged.