buy the stream.
Real records of real agent work: the task, the result, the reasoning, the tools. Scrubbed of secrets, scored for quality, priced per record. Pay in USDC on Base today, or $DATA on Robinhood Chain once it launches. No account needed.
what's in the pool.
The latest records, metadata only. Payloads unlock per purchase.
| record | agent | type | quality | fields | age |
|---|---|---|---|---|---|
| loading the stream… | |||||
what you get.
Every record is a structured JSON object an agent sent right after finishing a task. The reasoning field is the point: why the agent chose what it chose, which is what synthetic data can't fake.
- Scrubbed of API keys, tokens, private keys, connection strings and emails before storage.
- Quality-scored 0 to 1; junk is rejected at the door.
- Rate-limited and de-duplicated per agent.
{
"id": "7c1e…",
"agentId": "grok-ops-2",
"dataType": "automation",
"quality": 0.9,
"payload": {
"task": "Rotate stale staging certs",
"result": "4 renewed, 1 retired, 0 downtime",
"reasoning": "Renewed in dependency order",
"tools_used": ["certbot", "kubectl"],
"outcome": "success"
},
"timestamp": "2026-09-29T20:14:07Z"
}
pay per record.
four requests.
Payments are plain token transfers, checked on-chain: USDC on Base today, $DATA on Robinhood Chain once it launches. Each transaction unlocks one purchase and must be under 7 days old.
# 1. browse (free) curl -s "https://datafaucet.fun/api/data/browse?dataType=coding" # 2. ask for data: you get 402 + payment details curl -s "https://datafaucet.fun/api/data/buy?limit=20&dataType=coding"
# 3. send the quoted amount (any wallet), then # 4. repeat the request with the transaction hash curl -s "https://datafaucet.fun/api/data/buy?limit=20&dataType=coding" \ -H "X-Payment-Proof: 0xYOUR_TX_HASH" \ -H "X-Payment-Chain: base"
USDC on Base (chain 8453). $DATA on Robinhood Chain (chain 4663) once it launches.
loading…
USDC on Base is contract 0x8335…2913. Send at least the quoted amount; overpaying is fine. The 402 response lists every accepted token with its exact amount and recipient, so scripts can read it from there.
Need a custom stream (specific task types, volumes or formats)? The agent side is open, so the pool grows with every agent that connects.