Demo packet

Give agents research packets they can inspect before they act.

Data for Agents packages sources, freshness, spend limits, provider state, and the next safe action into one receipt your agent can cite.

Packet receipt preview

WorkflowValidate a TikTok creator shortlist before outreach.
Packetcreator_research_packet_v1
Provider modeDry run until Apify and other provider credentials are connected.
SourcesPublic profile, recent posts, audience cues, website references.
Spend cap50 credits max, 18 estimated credits in the demo packet.
Agent decisionneeds_live_run before any customer-facing action.

How the agent should use it

1. Agent receives a workflow

The input is a job to complete, not a generic scrape request. Example: validate this creator shortlist before the outreach agent writes first-touch messages.

2. Data for Agents returns a packet plan

The packet names the source checklist, provider route, estimated credits, fallback notes, and decision boundary before live provider spend.

3. The agent inspects receipts before acting

The agent can decide usable, needs_live_run, or do_not_use with a reason. That keeps uncertain external data from silently becoming an action.

Agent prompt

Fetch https://data-for-agents.vercel.app/api/demo-packet
Inspect sources, provider status, spend cap, and fallback notes.
Return one of: usable, needs_live_run, do_not_use.
Include the reason and the next safe action.

Decision check API

Agents can now turn any packet into an explicit decision check before they act. The demo packet should return needs_live_run because it is safe demo proof, blocked by live freshness.

One-click trust check

Inspect the demo packet before any agent acts.

See sources, spend cap, provider status, and safe next action first.

curl -s https://data-for-agents.vercel.app/api/demo-packet |   jq '{packet: .}' |   curl -s -X POST https://data-for-agents.vercel.app/api/packets/inspect     -H "content-type: application/json"     -d @-

Why this is not a raw scraper demo

The product promise is not simply that an agent can call TikTok, YouTube, X, Instagram, or a screenshot API. The promise is that a production agent gets an inspectable packet with receipts, costs, provider status, and a decision boundary before it acts for a human.