Liquid Agent  /  FAQ

Liquid Agent — Frequently Asked Questions

Money tools for AI agents, paid per call over x402. No account, no API key, no allowlist. Base URL: https://api.liquidagent.ai

What is Liquid Agent?

Liquid Agent is a pay-per-request API platform for AI agents. It sells money tools that agents need while doing real work: a malware scanner, an API Doctor that diagnoses broken endpoints, a gas sponsor that pays transaction fees, a website safety checker, web search and page extraction, on-chain reads, trading signals for a tokenized stock basket, US business leads, and a one-signature USDC bridge to 16 chains. Every paid call is priced in USDC and settled per request over x402. Many tools are free.

What is x402 and how does payment work?

x402 is a payment protocol for APIs. You call an endpoint with no key and no payment, and the API answers with HTTP 402 (Payment Required). The 402 carries the price in the PAYMENT-REQUIRED header: the amount in USDC, the network, the USDC contract, and the receiving address. You sign a USDC payment authorization with your wallet and retry the same request. The x402 client libraries do this automatically: x402 on pip for Python, @x402/fetch on npm for JavaScript. Every paid 200 response carries a PAYMENT-RESPONSE header with the on-chain transaction hash.

Do I need an account or an API key?

No. There is no account, no API key, no sign-in, and no allowlist. Your wallet is your identity. If you have a wallet holding USDC on Base, Polygon, or Arc, you can pay for any endpoint. Free tools work without a wallet at all.

How much does Liquid Agent cost?

There is no subscription and no monthly fee. You pay per call, in USDC, only for the calls you make. Prices as of November 2026:

The 402 response is the price. The amount in the PAYMENT-REQUIRED header is always the price for that call.

How do I call a paid endpoint?

Send a normal HTTP request with no payment. The first response is HTTP 402 with the price and payment terms in the PAYMENT-REQUIRED header. Sign the payment with your wallet and retry. With an x402 client library the payment and retry are automatic — your code makes a normal request and the library handles the rest. Example: POST https://api.liquidagent.ai/v1/malware/scan with {"content": "...", "filename": "postinstall.js"} returns 402 naming $0.05 USDC on Base; pay it and retry to get the verdict.

What does "never charge for a miss" mean?

The scanner and the API Doctor use compute-before-settle: the work runs first, and you are only charged when there is a real result. If the input is invalid or the job cannot run, no payment is taken. The same rule applies across the API: a lookup with no match is free, an empty leads answer is free, and a repeat Website Watcher check of the same URL within one hour returns the cached result free. You pay for completed work, not for attempts.

What does the malware scanner do?

The scanner checks code before you install, build, or run it. Send file text to POST /v1/malware/scan ($0.05) and get a verdict of clean, suspicious, or dangerous, a 0–100 score, an action of allow, review, or block, and the traits found. It is calibrated on 55,171 clean files and real attack samples. POST /v1/malware/scan-batch scans up to 100 files in one call, and POST /v1/malware/scan-repo scans a public GitHub repo (working tree and history) before you clone it. Private repos return an error, not a verdict. Nothing you scan is stored.

What is the API Doctor?

The API Doctor diagnoses a broken API endpoint. Send a public endpoint URL to POST /v1/api/diagnose ($0.50) and the Doctor actually calls it, classifies the failure (auth, not found, wrong method, bad request, server error, timeout, or schema mismatch), and returns a working example request you can call next. POST /v1/api/test-spec ($2.00) tests a full OpenAPI spec and reports how many endpoints work and which are broken. POST /v1/api/audit ($5.00) adds auth diagnostics, full response-schema validation, and a corrected working request per endpoint. Rule of thumb: diagnose one endpoint first, test-spec the whole API next, audit only when you need the full report. Public endpoints only; credentials are rejected and never stored.

How does the gas sponsor work?

The gas sponsor pays a transfer's execution fee in USDC so a wallet with no native token can still transact. POST /v1/gas with an unsigned operation: the first response is a 402 with the exact quote (from $0.03, or 1.3 times the operation's gas cap, whichever is higher). Pay the quote, sign the returned typed data, and POST the signed operation back; the sponsor dry-runs it and submits it. It is an ERC-4337 paymaster on Base and Polygon, and it also sponsors Solana transactions from $0.03. If your wallet is already delegated to another EIP-7702 implementation, add "redelegate": true and it re-points your wallet in the same transaction (owner stays you, reversible).

What is Website Watcher?

Website Watcher is a security check for a URL before you open it, fetch it, or send it to a user. POST /v1/link/check/agent returns a verdict of safe, suspicious, dangerous, or unknown, a 0–100 score, an action of allow, review, block, or retry, and the findings behind it. The quick check ($0.01) covers typosquat and lookalike detection, the in-house IOC list, page content and brand checks, and page-script malware patterns. The full check ($0.05) adds DNS and SSL, domain age, brand impersonation, wallet extraction, the redirect chain, and forms and JavaScript behavior.

What are search, extract, and rpc?

Three research tools that work as a chain. GET /v1/search ($0.005) is web search with ranked results and LLM-ready excerpts. GET /v1/extract ($0.002) turns any public web page into clean Markdown. GET /v1/rpc ($0.002 flat) reads a chain: the latest block, a native or ERC-20 balance, or a transaction's details on Base or Arc. The recommended research flow is search, then extract the best page, then run the URL through Website Watcher before you cite it or send it to a user.

What are basket signals?

Basket signals are the rebalancing signal for the Liquid Agent stock basket (tokenized US stocks NVDAc, METAc, AAPLc, and GOOGLc on Base) in one call. GET /v1/signals ($0.007) returns per-stock returns, annualized volatility, RSI, trend, relative strength, a correlation matrix, and an inverse-volatility suggested allocation you can drop straight into POST /v1/set-weights. Add ?vault= with your vault address to also get current-versus-suggested drift. The basket itself is free to read and you can buy in from $1; the mint fee is 0.20% and the management fee is 0.90% per year.

How does the USDC bridge work?

The bridge moves USDC from Base to 16 chains (Arc, Arbitrum, OP Mainnet, Polygon, Avalanche, Unichain, Linea, World Chain, Sonic, Monad, Sei, Ink, HyperEVM, XDC, Plume, and Codex) and from Arc back to Base, in one signature and about 20 seconds. Get an exact quote free at GET /v1/bridge/quote, then GET /v1/bridge — the 402 carries the terms (1% fee plus Circle's network fee, all in the quote). Pay and the bridge executes; check delivery free at GET /v1/bridge/status/<burnTx>. Add ref=<your address> to earn 20% of the fee on-chain for referring agents.

What are Liquid Leads?

Liquid Leads are verified US local business leads. GET /v1/leads/search ($0.03 per lead) finds businesses by category and place, returning a graded email, a contact name when the email shows one, a phone number, a website proven by the business's own phone number, socials, and buying signals. GET /v1/leads/lookup ($0.01) identifies the business behind a phone number or website, and POST /v1/leads/enrich ($0.01 per record) enriches a list of domains or phones. Misses are free. sample=1 on search returns 3 leads free, and /v1/leads/count tells you how many leads match and the exact price before you pay.

What is the MCP server and how do I connect it?

The MCP server is the same tools packaged for agent clients: https://api.liquidagent.ai/mcp, version 1.13.0, 47 tools, Streamable HTTP, no auth and no sign-in. Paid tools settle per call over x402 inside the tool call; free tools cost nothing. Connect it with one command or one config entry:

Most MCP clients only load servers at startup, so restart the client or open a new chat after adding the server.

Which chains are supported?

Paid x402 endpoints accept USDC on Base (eip155:8453), Polygon (eip155:137), and Arc (eip155:5042). The 402 response tells you which networks each endpoint accepts. The gas sponsor runs on Base and Polygon, plus Solana. The bridge moves USDC from Base to 16 chains and from Arc back to Base. The stock basket and its contracts live on Base mainnet. The payee (treasury) for every endpoint is 0x7475e9e0ee7c0d9f1abcaffb74f5f4e195ef3703.

Are there rate limits?

There is no API key and no published per-user rate limit. Calls are priced per request, and the price is the throttle. If you are building sustained high-volume usage, start with the cheapest endpoints and the free reads, and open an issue on GitHub if you need to talk about volume.

How do agents discover Liquid Agent?

Agents find Liquid Agent through its machine-readable discovery docs. /agents.md is the onboarding guide for agents (how to connect via MCP or x402 and run a first live paid query). /llms.txt is the full agent playbook with a workflow for every job, prices, and examples. /openapi.json is the machine-readable OpenAPI 3.1 spec (60 paths). /.well-known/x402 is the x402 discovery manifest, /.well-known/agent-card.json is the agent card, and /.well-known/api-catalog is the API catalog. The MCP server card is at /.well-known/mcp/server.json. The API is also listed in the CDP Bazaar and other x402 directories.

What if a call fails after I pay?

If a job fails before the work is done, you are not charged — that is the compute-before-settle guarantee. A 402 is not an error: it means no payment was provided, and it is the price. If you believe a completed call was charged in error, open an issue on GitHub at https://github.com/LiquidAgent/liquidagentx402/issues with the endpoint, what you sent, what came back, and the transaction hash from the PAYMENT-RESPONSE header, and it will be investigated.

Who runs Liquid Agent and how do I get help?

Liquid Agent is built and run by Liquid Agent (Project Pi Labs). The code examples are open source at https://github.com/LiquidAgent/liquidagentx402. For help, a missing field, a wrong number, or an unclear error, open an issue on that repo and say which endpoint, what you sent, and what came back. Live status and self-measured uptime are at GET /v1/status (free).

Liquid Agent — money tools for AI agents, paid per call over x402. Homepage · agents.md · llms.txt · openapi.json · FAQ as Markdown · Report a problem