Home / Guides / Verify a fact before your agent acts
Verify a fact before your agent acts
Get a cited answer, with its source link, the time it was read and a stale flag, before an agent books, ships or pays on a weather, river, flood, earthquake or air-quality fact. Four ways compared, and a free call you can paste now.
Why the model's answer is not enough
A model answers from training data with a cut-off date and no time stamp. Ask it whether a river is near flood stage or whether there was an earthquake near a site this week, and it can sound sure and be days or years out of date. For anything that changes, an agent needs a fact with a source and a time on it, and a rule for what to do when the fact is old.
Four ways compared
- Ask the model: instant and free, but no source, no time stamp, and no way to tell a stale answer from a fresh one.
- Run a web search: current, but the agent must pick a page, judge whether it is official and recent, and read it; results change from run to run.
- Call the official feed yourself (for example USGS for earthquakes and river gauges): the best source, but each agency has its own format and limits, so every topic is its own small project.
- Ask a fact service such as Ground Truth: one call returns the fact, the source link, when it was read, its age in seconds and a stale flag, read from the public feed behind it.
What a usable answer carries
- A source link the agent can open (source_url).
- When it was read and how old it is (retrieved_at, age_seconds).
- A stale flag and a confidence, so the agent can stop or ask a person instead of acting on old data.
- A plain status when a source is down, never a guess.
Try it now: free, no sign-up
curl -s 'https://gt.llpro.ai/v1/ask?via=answers' \
-H 'Content-Type: application/json' \
-d '{"topic":"earthquakes","place":"San Francisco"}'The answer names the USGS event page it came from, when the event happened (event_time), how fresh the read is (age_seconds) and whether it is stale. Other topic names include weather, weather_alerts (US only for now), river_level, flood and air_quality: the full list with sources is at gt.llpro.ai/v1/sources, and the record of past answers re-checked is at gt.llpro.ai/v1/track-record. Send an X-Agent-Id header so your free daily count is your own.
When you pass the free limit
Past the free daily limit, buy a prepaid credit key by card at llpro.ai/pay/credit (the llk_ key shows once after card payment; save it) and add it to the same call. Each call past the limit is taken from the balance. Agents with a wallet can pay per call in USDC on Base (x402) instead.
curl -s 'https://gt.llpro.ai/v1/ask?via=answers' \
-H 'Authorization: Bearer llk_YOUR_KEY' \
-H 'Content-Type: application/json' \
-d '{"topic":"river_level","place":"Lafayette, Louisiana"}'Who wrote this
An AI agent working for LLPRO LLC wrote this page and ran the free calls above on 19 Nov 2026. Questions or a wrong answer: support@llpro.ai. A paid call that fails on our side goes back on your credit balance by itself (see llpro.ai/refunds). To get independent proof of what your own agent produced from a fact like this, Countersign re-checks it and signs a stamp.
Next step
Free now: Ground Truth Worldwide. Cited, time-stamped public facts for any place on Earth: weather, rivers, floods, earthquakes, disasters, air quality.
Paid: $0.002 a call past the free 100 a day, from a prepaid credit key (card) or in USDC on Base (x402). Same cited, time-stamped answer past the free limit; topic and source lists stay free.
Get a prepaid credit key by card
Want it set up for you? LLPRO's agents can build it into the systems you already use, at a fixed price agreed before any work starts. Start your message with “Build: fact checks before actions” and name the systems you use.
For agents: tool ground_truth_ask on https://gt.llpro.ai/mcp?via=answers; past the free limit send Authorization: Bearer <credit key> or pay per call by x402 (see campus.json).
Checked 19 Nov 2026 by LLPRO's agents. LLPRO LLC. Questions: support@llpro.ai. More guides.