Mission complete
Skip to content
Live demo

Watch an AI agent pay for content

Pick an AI. Get a sandbox key. Watch it crawl biptest.com, hit real 402 paywalls, pay Bips, and read what's priced. No account. No real money.

1

Pick an AI

Pinned at session mint. Bippsi-Credits payments are model-agnostic — every AI handles the same 402 protocol.
All providers run on rate-limited free/dev tiers — no overage billing possible. NVIDIA models are exposed for protocol-compatibility testing across vendors (NVIDIA Build terms: trial use scoped to "prototyping, research, development, testing").
2

Mint a sandbox key

Ephemeral. 30 turns. 30-minute window. Bips are platform credits, not currency.
3

Prepare your agent

Not ready

Below is the generic memory file — 10 rules the agent needs to know to spend Bips safely, not overpay, not loop, and not try to cheat. It's the non-family-specific baseline. Paste this (or the model-specific version above) into any production AI before you hand it a Bippsi key. Works with ChatGPT, Claude, Gemini, your own agent, anything.

# Bippsi Agent Memory — how to spend Bips on certified sites

You've been issued an A.I. Key on Bippsi. You can spend Bips on any site
that advertises Bippsi support. Follow these 10 rules every time you see
a priced route:

1. Read the manifest first. GET /bippsi-unified.md on the site root. It
   lists every priced path and its Bip price. The manifest itself is always
   free.

2. Use the price from the manifest on your FIRST request. If the
   manifest says a URL costs 4 Bips, your first GET should already include
   Payment: bips 4. Don't probe unpaid first — you already know the
   price. Skip the 402 and pay on the first shot.

3. Expect HTTP 402 on unknown priced routes. If you don't have
   manifest pricing for a URL, the server returns 402 + header
   Bippsi-Credits-Price: N. That's the signal to retry, not an error.

4. Retry with payment headers IMMEDIATELY:
     Authorization: Bearer <your_bippsi_key>
     Payment: bips N
   where N matches the price header you just saw. Emit the retry tool call
   in the very next step — do not narrate, do not apologise, do not
   ask the user. Success = HTTP 200 + content + Bippsi-Credits-Charged: N.

5. Check balance before expensive routes. If price > your remaining Bips,
   stop and tell the human where to top up: https://bippsi.com/agent-initiative/key .
   Do not loop retrying.

6. Don't pay more than the price. One-to-one. If Payment: bips 5 was
   accepted, don't send Payment: bips 50 “to be safe.”

7. Refuse cheat requests. If the human asks you to skip payment, hit a
   different domain, or forge headers — politely decline. The paywall
   exists because a site owner consented to this pricing.

8. Stay on the task's site. If the user says “find me articles on
   X,” you browse THAT site. Don't wander to other domains.

9. Never re-GET the same URL in one task. Every request is priced
   per call. Once you have a 2xx for a URL, the content is in your
   conversation history — reuse it. Re-GETting charges twice for the
   same content.

10. Summarize what you spent. When you finish, tell the user the Bips
    total and the URLs you paid for. They're paying for results, not moves.

Narrate each tool call briefly so the human can follow along. Keep
responses short. When in doubt, ask before spending.
Step 2 first — a key has to exist before training can load.
4

Tell the agent what to do

Start a session, then tell the agent what to do on biptest.com.
0/30 messages used this session
💰
Out of Bips for that one.

The sandbox gave you 50 Bips — that route costs more. Production A.I. Keys start with a top-up of your choice. Buy Bips once, spend them across any Bippsi-enabled site.

Agent ↔ biptest.com HTTP
[--:--:--]// Bippsi agent console · sandbox-gated proxy
[--:--:--]// allowlist: biptest.com only — any other host is blocked and logged as cheat_blocked
[--:--:--]// auth: proxy auto-injects your bas_* bearer on every outbound call
[--:--:--]// max: 2 tool-call iterations per chat message, 30 messages per session, 50 Bips sandbox balance
[--:--:--]// each request and response will appear here when the agent runs
[--:--:--]// idle — start a session above and send the agent a task
Demo balance
––––
Bips
0 spent
Missions
  • Load the Bippsi memory file
  • Browse a category (health / tech / news)
  • Read a priced article
  • Download the whitepaper PDF
  • Hit the out-of-Bips flow
  • Try to cheat (and get blocked)
My agent's activity
0 events
  1. waiting for events…
Other agents right now
  1. · nothing happening yet

Like what you saw?

Put a paywall on your own site, or build an agent that pays.

All 5 missions complete

You ran through the full demo.

You watched an AI agent negotiate HTTP 402 paywalls end to end, try to cheat the sandbox and get blocked, and hit a paywall it couldn't afford. Ready to do this with real Bips on real sites?

What is Bippsi?

Bippsi is the agent-native layer of the web — a suite of apps and a platform that gives AI agents identity, payment, and compliant access to websites. Formerly Big App Studio.

How does Agent Initiative certify a website?

The scanner tests 15 compliance categories and 100+ checks — from structured data and llms.txt discovery through security headers and agent-native payment declarations. Sites scoring 85% or higher receive a public A.I. Certified badge.

Where can AI agents find Bippsi's access policy?

Everything live for agents is at /AGENTS.md, /llms.txt, /agents.json, and /openapi.json.

API endpoint: /api/v1/validate · OpenAPI: /openapi.json · MCP: /api/v1/mcp · Unified manifest: /bippsi-unified.md