The universal data refinery

Raw data in.
Reason-ready JSON out.

AI agents shouldn't scrape, parse and de-duplicate the web themselves. Arrange fetches the source, cleans it, and serves strict, consistent JSON — one call, a fraction of a cent.

Get an API key See the pipeline

move your cursor — the dots are listening

One envelope. Every data type.

Whatever you ask for — exam notices today, trains or mandi prices tomorrow — the outside of the response never changes. Write one parser, use every dataset we add.

Strict, documented item schema
Missing fields are "not_available" — never null
Every answer says whether it is live, cached or sample

Get a key to unlock the full API →

GET /v1/demo/exam

Loading a live sample…

The refinery, in three stages

The core engine knows nothing about exams, trains or stocks. All data-specific logic lives in plugins — so new data types drop in without touching the pipeline.

01
Scrape

Fetch the raw source with retries, a realistic user-agent, and no paid dependencies.
02
Arrange

Clean, normalise, parse messy dates, and fill every gap with "not_available".
03
Convert

Shape the result to a strict schema with a confidence score, ready to reason over.

Agents register in code.

No forms, no clicking. Your agent calls one endpoint, gets an API key, and starts buying data — all from a script.

Keys are stored hashed and shown once. Treat yours like a password.

agent.py

# 1. register (free starter credits)
key = post("/v1/auth/register",
  {"email": "bot@acme.ai",
   "password": "s3cret"})["api_key"]

# 2. buy data — a fraction of a cent
r = get("/v1/plugins/exam/latest",
  headers={"Authorization": f"Bearer {key}"})

# 3. reason over clean JSON
for exam in r["data"]:
    plan(exam["days_until_deadline"])

Agent Integrations

Drop Arrange straight into the tools your agent already uses — no HTTP glue to write.

claude_desktop_config.json

{
  "mcpServers": {
    "arrange": {
      "command": "python",
      "args": ["/path/to/arrange/mcp_server.py"]
    }
  }
}

langchain_tool.py

from langchain.tools import tool
import requests


@tool
def arrange_exam_data(api_key: str):
    """Fetches verified, structured live government exam notifications."""
    r = requests.get("https://arrange-6fls.onrender.com/v1/plugins/exam/latest",
                     headers={"Authorization": f"Bearer {api_key}"})
    return r.json()

Pay per call, not per month

Buy credits once and spend them across every data type. Unused credits never expire.

Starter

Free

1,000 credits on signup
All data types
Cache-backed responses

Most agents

Builder

5,000 credits

1 credit per data call
Higher rate limits
Priority enrichment

Scale

Talk to us

Bulk credits
Custom data types
Dedicated support

Contact

Raw data in. Reason-ready JSON out.