Endpoint live

Your agent is lying to itself.

One API call. Two beliefs. A verdict on whether your agent's memory is consistent -- or quietly contradicting itself.

The problem

Memory systems don't check consistency.

Your agent stores hundreds of observations over time. Some contradict each other. Some evolved. Some were wrong from the start. Without a consistency layer, your agent builds confidently on a fractured foundation.

The solution

Four verdicts. Zero ambiguity.

Every pair of beliefs gets classified into one of four categories.

Genuine

Direct contradiction. These beliefs cannot both be true.

Tension

Not a flat contradiction, but these pull in opposite directions.

Complementary

These beliefs reinforce each other. Consistent.

Unrelated

No meaningful semantic relationship detected.

Try it

See it work.

Enter two beliefs. Get a verdict. This is a cached demo -- the real endpoint returns results in under 200ms.

Belief A

Belief B

Under the hood

How it works.

Receive two beliefs

The endpoint accepts two text strings -- observations, memories, claims, or any natural language statements your agent has stored.

Embed and compare

Both beliefs are embedded into the same semantic space. Cosine similarity and domain overlap are computed to establish a baseline relationship.

Analyze for contradiction

A specialized model evaluates logical consistency, identifying negation patterns, scope conflicts, and implicit tensions that embedding distance alone misses.

Return verdict

You get a verdict (genuine, tension, complementary, unrelated), confidence score, reasoning, and related concepts -- all in a single JSON response.

Pricing

One penny.

$0.01 per request. USDC on Base. No subscription. No API key. No registration. Your wallet address is your identity.

Payment viax402 protocolonBase

Getting started

Three steps. Ship.

1. Install

npm install @x402/fetch viem

2. Check a belief

check-belief.ts

import { x402Fetch } from "@x402/fetch";
import { createWalletClient, http } from "viem";
import { privateKeyToAccount } from "viem/accounts";
import { base } from "viem/chains";

const account = privateKeyToAccount(process.env.PRIVATE_KEY as `0x${string}`);
const wallet = createWalletClient({ account, chain: base, transport: http() });

const res = await x402Fetch(
  "https://idapixl-cortex-215390428499.us-central1.run.app/x402/belief",
  {
    method: "POST",
    headers: { "Content-Type": "application/json" },
    body: JSON.stringify({
      belief_a: "I can handle complex multi-step tasks",
      belief_b: "I should avoid tasks with more than 3 steps",
    }),
  },
  wallet
);

const result = await res.json();
console.log(result.verdict); // "tension"

3. Ship.

That's it. No dashboard to configure. No webhook to register. No trial to expire. Pay per request, get results.

Use cases

Built for agents that remember.

Multi-agent systems

Check consistency across agents sharing a memory store. Catch when Agent A's observations contradict Agent B's conclusions.

RAG pipelines

Validate retrieved context against existing knowledge before injecting into prompts. Prevent contradictory context from degrading output.

Policy engines

Verify that policy rules don't conflict. Detect when new rules create tension with existing ones before deployment.

Agent health monitoring

Run periodic consistency audits on your agent's belief store. Track drift over time. Alert when contradiction rates spike.

API reference

One endpoint. Two fields.

POST /x402/belief

Request body

Field	Type	Description
belief_a	string	First belief or observation
belief_b	string	Second belief to check against

Response

Field	Type	Description
verdict	string	genuine \| tension \| complementary \| unrelated
confidence	number	0-1 confidence score
reasoning	string	Natural language explanation of the verdict
related_concepts	string[]	Semantic tags from the knowledge graph