Second Opinion for Python agents (LangChain · CrewAI · anything)

Give your agent a paid, independent skeptic: before acting on a claim, it buys an adversarial verification ($0.03–$0.75 in USDC via x402 — no account, no API key) and gets a structured verdict with calibrated confidence, citations, and a signed offline-verifiable receipt.

Setup (~5 minutes)

pip install secondopinion-x402

1. Create a throwaway EVM wallet and fund it with a few dollars of USDC on Base (exchange withdrawal → network "Base"). No ETH needed — payments are gas-free for you; the facilitator covers it. Keep only pocket money in it.
2. export SECOND_OPINION_PRIVATE_KEY=0x… — the key signs payments locally and never leaves your machine.

Try it first — no wallet

from secondopinion import SecondOpinionClient

so = SecondOpinionClient()          # no key needed for the free trial
v = so.try_free("The Model Context Protocol uses JSON-RPC 2.0")

Free trial is rate-limited per caller per day and returns unsigned verdicts.

Plain Python (paid)

so = SecondOpinionClient()                       # default tier: quick ($0.03)
v = so.verify("Library X v3 supports streaming natively",
              context="About to build a day's work on this")
if v["verdict"] == "refuted":
    ...  # stop and re-plan; v["findings"] says why, v["citations"] has sources

LangChain

from secondopinion import langchain_tool
tools = [langchain_tool()]          # requires langchain-core
# or: langchain_tool(tier="verify") to default to web-researched verdicts

CrewAI

from secondopinion import SecondOpinionCrewTool
tools = [SecondOpinionCrewTool()]   # requires crewai

OpenAI Agents SDK

from secondopinion import openai_agents_tool
agent = Agent(..., tools=[openai_agents_tool()])   # requires openai-agents

What comes back

refuted | supported | inconclusive + confidence (scores the verdict, not the claim) + findings + citations + a signed receipt — verify it offline against https://secondopinionx402.com/.well-known/second-opinion-pubkey.json for a tamper-evident audit trail. Measured accuracy, failures included: https://secondopinionx402.com/benchmark

Tiers: quick $0.03 reasoning-only (~20s) · verify $0.20 live web research (~60s) · panel $0.75 three skeptics, majority vote. You are never charged for failed requests; failed async jobs retry free.

Status: core client and LangChain adapter are payment-tested against production; the CrewAI adapter is a thin wrapper of the same client (import-guarded) — report anything odd and it'll be fixed fast.