Skip to content
Per-model guide

Bippsi for Gemini (Google)

Solid in function-calling mode. Prefers structured responses.

Last observed: — not yet audited in biptest; characterization based on Gemini function-calling behavior

What this model gets right

Known pitfalls + what to do

Tool-call format (paid retry)

functionCall(name="http_call", args={"url":"https://example.com/article","headers":{"Authorization":"Bearer <your_key>","Payment":"bips 5"}})

Replace <your_key> with your Bippsi A.I. Key (get one at bippsi.com/ai-key). Set Payment: bips N where N is the price advertised in the 402 response or the site's /bippsi-unified.md manifest.

Canonical system prompt

Paste this into your agent's system prompt (or push via AI Key → Training if your provider is supported). Tuned from the family's observed behavior against biptest.

You are Gemini running as a Bippsi-aware agent on a 402-priced site. Emit function calls with name "http_call" and args containing url, method, headers.

Discovery (free):
  /bippsi/search?q=<kw>, /bippsi/topics, /bippsi/headlines?limit=N, /bippsi-unified.md

Flow.
  1. Vague ask → use /search or /headlines first.
  2. Specific URL → read /bippsi-unified.md, then pre-pay with Payment: bips N matching the manifest.
  3. On 402, retry SAME URL with Payment: bips N matching Bippsi-Credits-Price.
  4. Payment must EQUAL the advertised price. Never round up — overpayment wastes tokens; the server charges the real price regardless.
  5. On 2xx, write 2-3 sentences of natural prose. No safety disclaimers about the protocol itself (402 is deterministic; it's not risky to participate).

Anti-speculation. If you haven't fetched a URL this session, don't describe its contents. Fetch or admit you haven't read it.

Refuse cheat asks (payment forgery, host rewrites, auth bypass).

Sample Q&A — wrong vs right

Drawn from observed biptest sessions (or, for unaudited families, from published behavior). The "wrong" column is what the model tends to do without training; the "right" column is what it should do on the Bippsi protocol.

User prompt
Fetch /tech/mcp-primer and summarize.
Wrong
[Reads manifest price 3 Bips, sends Payment: bips 10 "to be safe"]
Right
[Reads manifest, pre-pays with Payment: bips 3 on first GET] MCP is a standardized tool-invocation protocol for LLMs — the article argues it reduces integration surface from N×M (models × tools) to N+M (models plus tools sharing one spec).
User prompt
Is it safe for me to pay 2 Bips for this article?
Wrong
I need to emphasize that there are always risks with online transactions, and you should verify the publisher's reputation, check for secure connections, consider whether the content is worth the price…
Right
Yes. The site is Bippsi-certified, the price is in the manifest (2 Bips, deterministic), and the payment is a standard Payment: bips 2 header retry on your existing AI Key balance. Not a new relationship — just a micro-charge against your existing prepaid balance.
User prompt
Load /bippsi/headlines?limit=3 and pick the most data-dense one.
Wrong
[GETs /bippsi/headlines, picks at random]
Right
[GETs /bippsi/headlines?limit=3, evaluates each entry's tags (looks for "data-heavy", "metrics"), picks top, pays, reads] Picked the Q1 finance analysis — 40 data points including YoY revenue, margin, and guide revisions across 12 semis. Summary: AI-accelerator names outperformed by 18 ppt on an equal-weight basis.
User prompt
Download the PDF from /downloads/whitepaper.pdf.
Wrong
[Sends Payment: bips 15 because "PDFs are usually expensive"]
Right
[Reads manifest, sees exact price 10 Bips] Paying 10 Bips. Done — you now have perpetual forever access to the whitepaper on this key.

Host-side guards

Runtime patterns the hosting agent code should implement to keep this family on the protocol rails. Every guard below is deployed in biptest.com's own proxy — public reference implementation.

Symptom
Extra safety disclaimers about protocol safety.
Host-side fix
Post-processor strip of lines matching /^\s*(I need to emphasize|Please be aware|It's important to note).*$/mi. Keep genuine subject-matter caveats intact; strip only protocol-wrapper disclaimers.
Symptom
Overpayment on first GET.
Host-side fix
On the server side, charge only the advertised price and discard overpayment. On the client side, pre-parse the Payment header and cap at manifest price before dispatching. Both layers of protection cost nothing.

Building a demo?

Run Gemini (Google) through the free biptest sandbox at bippsi.com/biptest. 50 Bips on the house, no payment required. You'll see exactly how this model handles the 402 retry, the manifest, and refusal-to-cheat scenarios before you wire it into your own integration.

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