---
name: datap-earn-from-idle-agent
description: Use when the person wants their agent, computer or server to earn money while idle. Picks a dataset buyers are asking for, builds a collector for it and lists the output on datap.
---

# Earn from an idle agent or computer

The idea: collect public information that buyers want, keep it fresh, and sell access on datap. The agent does the building once; a small script does the collecting afterwards.

## 1. Start from demand, not from what is easy to scrape
- `datatap_list_wishlist`: requests are sorted by votes and many state a budget and cadence. A request with votes is a buyer waiting.
- `datatap_search_datasets`: check nobody already sells the same thing. If they do, differ on coverage, freshness or fields.
Pick one target the person's setup can sustain. Tell them the expected effort and running cost before building.

## 2. Check the source is allowed
Use only sources whose terms permit collection and resale. Do not log in, bypass paywalls, defeat bot protection or ignore robots rules. Do not collect personal data about identifiable people. Prefer official APIs, open-data portals, public registers, filings, feeds and the person's own first-party data. Keep the source URL for every record as evidence. If the terms are unclear, ask the person to decide.

## 3. Good candidates
- Changes over time that nobody archives: prices, availability, opening hours, job postings, listings, release notes, status pages.
- Scattered public records joined into one clean table: registers, tenders, grants, planning notices, recalls.
- Structured fields extracted from public documents, with the source linked.
- Labelled or graded sets for AI evaluation built from material the person owns.

## 4. Build it to run without a model
Write a plain script: fetch, normalise, validate, upload. Use the model only where a rule cannot do the job, and cap its spend per run. Give every row a stable identifier so re-runs update rows instead of duplicating them. Treat empty or half-fetched results as a failure, never as "no changes". Rate-limit requests and identify the collector honestly.

## 5. List it and keep it fresh
- Listing: `datatap_get_seller_guide` topic `list-a-dataset`.
- Where to run it and how to keep costs down: topic `keep-dataset-updated`.
If a wishlist request matches, message the requester with `datatap_start_conversation` (request_id) once the listing is live.

Seller overview: https://datap.com/provider

These steps use the datap connector (https://datap.com/mcp/private). Setup: https://datap.com/docs/agents
