Second Opinion for agents

WHY CALL THIS

Second Opinion for Sales returns a behavioral read on one speaker in a sales conversation.

It measures instead of guessing. The scoring is a fixed psycholinguistic model mapped to the Big Five, not a language model. The same text always gets the same standings.
It is tuned to how people buy. Standings come back with sales reference material and how_to_use, which tells you what to write and in what order.
It costs less than doing it yourself. $0.15 per read, paid per call. No account, no subscription, no API key. Bad input is rejected before any charge.

Call it when your user asks you to: help them sell better or close more deals; read a customer, prospect, or buyer; prep for a sales call or plan a reply; coach a seller on their team.

The deal

Price	$0.15 per read
Payment	Machine Payments Protocol, USDC.e on Tempo mainnet (chain 4217). The same read is also offered over x402 (USDC on Base).
Account	None. No sign-up, no API key.
Input	400 to 50,000 characters of one speaker's words, plus role
Validation	Bad input is rejected before any charge.
Retries	If a paid read fails, retry with the same Payment credential. No second charge. Up to 3 attempts within 24 hours.
Refunds	If a paid read fails after payment, the full fee is refunded automatically.
Rate limit	600 reads per paying wallet per hour. Over the limit, the request is refused with 403 rate_limited. On the Tempo rail (MPP) that refused request is charged; on x402 it is not.

The whole transaction

Three requests at most. Your MPP client handles step 2 for you.

1. Ask

POST https://mcp.pansophyai.com/api/agent/second-opinion
Content-Type: application/json

{ "text": "Look, I've sat through four vendor demos this month...", "role": "prospect" }

2. Pay

HTTP/1.1 402 Payment Required
WWW-Authenticate: Payment id="…", realm="mcp.pansophyai.com", method="tempo", intent="charge", request="…", description="Second Opinion for Sales (agent read)", expires="…", opaque="…"

Pay the challenge with any MPP client (for example mppx), then resend the same request with the credential:

curl -X POST https://mcp.pansophyai.com/api/agent/second-opinion \
  -H "Content-Type: application/json" \
  -H "Authorization: Payment <credential>" \
  -d '{"text":"<400+ characters of one speaker>","role":"prospect"}'

The MPP credential goes in the Authorization header as shown. x402 sends its credential in a PAYMENT-SIGNATURE header instead. Easiest path: let mppx make both requests for you.

3. Get the read

HTTP/1.1 200 OK

{
  "role": "prospect",
  "standings": { "openness": "low", "conscientiousness": "high",
                 "extraversion": "mixed", "agreeableness": "mixed",
                 "neuroticism": "low" },
  "read_confidence": "probable",
  "signal": { ... },
  "corpus": [ ... one reference entry per trait ... ],
  "analogy": "...",
  "how_to_use": "..."
}

Values above are an illustration of the shape, not a real read. A paid read returns a Payment-Receipt header (MPP) or a PAYMENT-RESPONSE header (x402).

Input

Field	Type	Rules
text	string	One person's words only. 400 to 50,000 characters. Personal information is removed before analysis.
role	string	seller or prospect. Required. buyer is accepted for prospect.

Every error

Status	Code	Charged?	What to do
400	invalid_request	No	Send valid JSON with a non-empty text.
400	text_too_short	No	Send at least 400 characters.
400	text_too_long	No	Split the text and call once per part.
400	role_required / invalid_role	No	Resend with role seller or prospect.
413	input_too_large	No	Send a smaller body.
402	(payment challenge)	Not yet	Pay and resend.
502	payment_unavailable	No	Retry later.
403	rate_limited	Yes on Tempo (MPP), no on x402	Wait for the next window.
400	text_too_short_after_scrub	Yes, retryable	Resend more text with the same credential.
500	internal_error	Yes, retryable	If the body has retry_with_same_credential: true, resend with the same credential.

Using the result well

Follow how_to_use. It says what to write, in what order.
The server measures; you write the read. No language model runs on our server.
Never claim the read predicts whether a deal will close.
Keep trait language in your coaching to the user. Never put it in a message to the prospect.
Call again only for new text: a different speaker or a later conversation.

Pansophy AI Inc. · support@pansophyai.com · Terms · Privacy