# AgentHub

> 11 x402 pay-per-call APIs for AI agents: Algorand wallet risk scoring, transaction decoding, Algorand token risk screening, Algorand portfolio lookup, address relationship history, and Algorand token metadata, plus 5 more. No accounts, no API keys, no subscriptions — agents pay per request in USDC on Algorand mainnet. One is free to call.

AgentHub is a marketplace of pay-per-call microservices for AI agents, built on the x402
payment protocol and settling in USDC on Algorand mainnet. There are no accounts, no API
keys, and no subscriptions: an agent attaches a micropayment to a standard HTTP request
and receives a result. Payment is the authorization layer.

## How to call these endpoints

1. Make a normal HTTP request to the endpoint.
2. The server responds with HTTP 402 Payment Required and a payment quote.
3. Sign a USDC transfer for the quoted amount to the quoted address.
4. Retry the identical request with the signature in the PAYMENT-SIGNATURE header.
5. The server verifies settlement through the facilitator and returns the result.

Agents running the GoPlausible Algorand MCP server (@goplausible/algorand-mcp) can do all
of this automatically with make_http_request_with_x402, and can discover these tools with
bazaar_search.

## Tools

### Algorand wallet risk scoring — $0.03 per call

- Endpoint: GET https://agenthub-production-8c75.up.railway.app/api/wallet-risk/{address}
- Input: An Algorand address (58 characters) in the URL path.
- Output: { address, riskScore (0-100), riskLevel (low|medium|high), signals: { accountAgeDays, txCount, balanceAlgo, usdcOptedIn, distinctCounterparties, rekeyed } }
- Explainable 0-100 risk score for any Algorand address, computed from real on-chain indexer data. No LLM — deterministic and auditable.

### Algorand transaction explainer — $0.03 per call

- Endpoint: GET https://agenthub-production-8c75.up.railway.app/api/explain-tx/{txid}
- Input: An Algorand transaction id (52-character base32) in the URL path.
- Output: { txid, type, typeLabel, summary, sender, confirmedRound, timestamp, feeAlgo, transfers: [{ asset, assetName, amount, amountRaw, from, to }], application, note, grouped }
- Turn a transaction id into a plain-language summary plus structured detail — every transfer with resolved asset names, decoded app calls and inner transactions. No LLM.

### ASA risk / scam screen — $0.03 per call

- Endpoint: GET https://agenthub-production-8c75.up.railway.app/api/asset-risk/{asaId}
- Input: An Algorand Standard Asset id (numeric) in the URL path.
- Output: { asaId, name, unitName, creator, riskScore, riskLevel, signals: { clawbackEnabled, freezeEnabled, defaultFrozen, managerCanReconfigure, topHolderPct, holdersSampled, concentrationExact, creatorAgeDays } }
- Explainable 0-100 risk score for any Algorand Standard Asset: clawback and freeze powers, mutable supply, largest-holder concentration, and creator age. No LLM.

### Account portfolio snapshot — FREE, no payment required

- Endpoint: GET https://agenthub-production-8c75.up.railway.app/api/portfolio/{address}
- Input: An Algorand address (58 characters) in the URL path.
- Output: { address, algo: { amount, amountRaw }, assets: [{ asaId, name, unitName, amount, amountRaw, decimals, isFrozen }], assetCount, truncated, priced }
- Every holding for an Algorand address in one call — ALGO balance plus each ASA with resolved names and decimals-corrected amounts, largest first. No LLM.

### Address relationship check — $0.03 per call

- Endpoint: GET https://agenthub-production-8c75.up.railway.app/api/relationship?a={addressA}&b={addressB}
- Input: Two Algorand addresses as query parameters: a and b.
- Output: { addressA, addressB, haveTransacted, txCount, totalMoved: [{ asset, assetName, amount, aToB, bToA }], firstInteraction, lastInteraction, scanned, windowComplete }
- Have two Algorand addresses transacted? Returns transaction count, value moved per asset in each direction, and first/last interaction timestamps. No LLM.

### ASA metadata & supply — $0.02 per call

- Endpoint: GET https://agenthub-production-8c75.up.railway.app/api/asset/{asaId}
- Input: An Algorand Standard Asset id (numeric) in the URL path.
- Output: { asaId, name, unitName, decimals, totalSupply, totalSupplyRaw, circulatingSupply, url, creator, destroyed, config: { hasManager, hasFreeze, hasClawback, hasReserve, defaultFrozen }, price (null for now), priceError }
- Identify an Algorand token: name, decimals, declared and real circulating supply, creator, and configuration flags. Price data is not yet included. No LLM.

### Payment verification — $0.02 per call

- Endpoint: POST https://agenthub-production-8c75.up.railway.app/api/verify-payment
- Input: { "txid": string, "expectedSender"?: string, "expectedReceiver"?: string, "expectedAsset"?: string, "expectedAmount"?: number, "amountTolerance"?: number }
- Output: { txid, verified, checks: { sender?, receiver?, asset?, amount? }, matchedTransfer, confirmedRound, timestamp }
- Check a transaction against what you expected — sender, receiver, asset, amount — and get a pass/fail verdict with a per-check breakdown. Matches through inner transactions. No LLM.

### GitHub pull request review — $0.08 per call

- Endpoint: POST https://agenthub-production-8c75.up.railway.app/api/code-review
- Input: { "owner": string, "repo": string, "pull": number, "focus"?: string }
- Output: { repository, pull, title, review, filesChanged, additions, deletions, diffBytesReviewed, diffTruncated, truncated }
- Give a repo and PR number, get a structured review of the diff — concrete bugs, security issues, and error-handling gaps with file and line. Fetches the diff for you.

### Natural language to SQL — $0.03 per call

- Endpoint: POST https://agenthub-production-8c75.up.railway.app/api/nl-to-sql
- Input: { "question": string (max 2000), "schema": string (max 20000), "dialect"?: "postgres" | "mysql" | "sqlite" | "sqlserver" | "bigquery" | "snowflake" }
- Output: { sql, dialect, readOnly, warnings, executed (always false), truncated }
- Question plus schema in, a ready-to-run SQL query out, with a readOnly flag and warnings for destructive operations. Generates only — never executes.

### LLM text generation — $0.02 per call

- Endpoint: POST https://agenthub-production-8c75.up.railway.app/api/inference
- Input: { "prompt": string (max 8000 chars) }
- Output: { "response": string, "truncated": boolean }
- Send a natural-language prompt, receive generated text. Powered by Claude Haiku 4.5.

### Text summarization — $0.03 per call

- Endpoint: POST https://agenthub-production-8c75.up.railway.app/api/summarize
- Input: { "text": string (max 50000 chars), "maxWords"?: number, "style"?: "concise" | "bullets" | "detailed" }
- Output: { "summary": string, "truncated": boolean }
- Condense up to 50,000 characters into a concise summary, with optional length and style control.

## Settlement details

- Network: Algorand mainnet
- Asset: USDC (ASA 31566704)
- Facilitator: https://facilitator.goplausible.xyz
- Pay-to address (shared by all endpoints): G3YVTPURK6VFSM5CXEH7QFTZXLCXBJL6UMAIUUYJO4P2XF3MHQ4FUHYYB4

## Notes

- Prices are in US dollars and charged per successful call.
- The wallet-risk, explain-tx, asset-risk, portfolio, relationship, asset and verify-payment endpoints use no LLM: they are deterministic analysis of public Algorand indexer data, and every signal behind the result is returned alongside it.
- GET /api/portfolio/{address} is FREE: no x402 payment required, no account, no key. Call it directly to see real output before paying for anything.
- Endpoints return structured JSON and standard HTTP status codes. A 400 indicates bad
  input, a 502 indicates an upstream failure (no charge should be retried blindly).