Loop gate

Should the loop run another iteration? Send POST /api/loop-gate with the required fields goal, lastOutput and iteration and pay $0.001 per call over x402 or MPP (there is no free tier). It returns the result.

Send the goal, the iteration that just finished (lastOutput, iteration), and optionally maxIterations, a budget { spentUsd, maxUsd }, the history of earlier iterations (up to 20 of { iteration, summary }, oldest first) and constraints; get back decision (continue, stop or escalate), reason, confidence (0 to 1), the checks that produced it (goalMet, budgetExceeded, iterationsExhausted, repeating, stuck, constraintBreached) and nextStep (iterate, change_approach, narrow_scope, verify_result, hand_off). The budget and iteration checks run in code before anything else: when either trips the answer is stop, no model is consulted, goalMet is null and model is null, at the same price. An output that repeats an earlier summary (repeating) and three near-identical outputs in a row (stuck) are found lexically first. Otherwise a typed judgment model (the /v1/judge backends) answers whether the goal is met, whether the loop is stuck, whether a constraint is breached and what to do next; model names which served. Model-backed, not deterministic, except the budget and iteration checks, which are computed in code before any model call.

Try in playground →

Parameters

Example request

curl -i -X POST https://agent402.tools/api/loop-gate \
  -H "Content-Type: application/json" \
  -d '{"goal":"Find three peer-reviewed studies on sleep and memory consolidation published since 2020, with DOIs.","lastOutput":"Found two studies with DOIs (Nature Neuroscience 2021, Sleep 2023). A third search of PubMed returned review articles only.","iteration":4,"maxIterations":6,"budget":{"spentUsd":0.52,"maxUsd":0.5},"history":[{"iteration":2,"summary":"One study found (Nature Neuroscience 2021)."},{"iteration":3,"summary":"Second study found (Sleep 2023); PubMed search for a third in progress."}]}'

Without payment this returns HTTP 402 Payment Required with the exact price for loop-gate; any x402 v2 or MPP client pays it and retries.

Example response

{}

From an MCP client

catalog.call {
  "slug": "loop-gate",
  "params": {
    "goal": "Find three peer-reviewed studies on sleep and memory consolidation published since 2020, with DOIs.",
    "lastOutput": "Found two studies with DOIs (Nature Neuroscience 2021, Sleep 2023). A third search of PubMed returned review articles only.",
    "iteration": 4,
    "maxIterations": 6,
    "budget": {
      "spentUsd": 0.52,
      "maxUsd": 0.5
    },
    "history": [
      {
        "iteration": 2,
        "summary": "One study found (Nature Neuroscience 2021)."
      },
      {
        "iteration": 3,
        "summary": "Second study found (Sleep 2023); PubMed search for a third in progress."
      }
    ]
  }
}

The hosted connector at https://agent402.tools/mcp needs a payment for loop-gate; the stdio package pays it from a wallet or from an AGENT402_CREDITS_KEY issued earlier. Local install: npx -y agent402-mcp.

Errors and behavior

• goal, lastOutput and iteration are required. An input the tool rejects returns an HTTP 4xx whose body carries error, tool, expected, required and example, so the caller can correct it.
• A paid call that ends in any status of 400 or above is not charged over x402, MPP or a prepaid credits key: settlement is canceled when the tool fails. The exception is a Tempo push credential, a transfer the buyer sent before the call: it settles before the tool runs, so if the tool then fails the payment is recorded as a refund owed to the paying wallet.
• Wallet-only: this tool runs a model, so it has no proof-of-work tier. A prepaid card-credits key issued earlier (Authorization: Bearer a402_...) also pays it.
• Model-backed: the answer is generated by a model, so the same input can produce a different result.
• A GET or HEAD to /api/loop-gate returns the same 402 quote, so the price can be read without a body.
• An Idempotency-Key header makes a retried paid call replay the first 200 instead of charging again (an answer larger than 1 MB is not replayed).

Paid call (JavaScript agent)

import { wrapFetchWithPayment } from "@x402/fetch";
import { x402Client } from "@x402/core/client";
import { registerExactEvmScheme } from "@x402/evm/exact/client";
import { privateKeyToAccount } from "viem/accounts";

const client = new x402Client();
client.setSpendControls?.({ maxAmountPerPayment: "$0.001" }); // refuse any quote above this tool's price
registerExactEvmScheme(client, { signer: privateKeyToAccount(KEY) });
const payFetch = wrapFetchWithPayment(fetch, client);

const res = await payFetch("https://agent402.tools/api/loop-gate", {
  method: "POST",
  headers: { "Content-Type": "application/json" },
  body: JSON.stringify({
    "goal": "Find three peer-reviewed studies on sleep and memory consolidation published since 2020, with DOIs.",
    "lastOutput": "Found two studies with DOIs (Nature Neuroscience 2021, Sleep 2023). A third search of PubMed returned review articles only.",
    "iteration": 4,
    "maxIterations": 6,
    "budget": {
      "spentUsd": 0.52,
      "maxUsd": 0.5
    },
    "history": [
      {
        "iteration": 2,
        "summary": "One study found (Nature Neuroscience 2021)."
      },
      {
        "iteration": 3,
        "summary": "Second study found (Sleep 2023); PubMed search for a third in progress."
      }
    ]
  }),
});

Related tools