One call. Multiple risk advisors.

Built for agents to inspect.

RugShield's own checks are Python rules over Solana data, not LLM calls. We use Helius for token data and transaction simulation, and GMGN for deployer statistics when available. The panel calls selected independent advisor APIs in parallel; their methods are independent of ours.

HTTP quickstart: inspect an unpaid HTTP 402 quote using the wrapped SOL mint. This direct call includes every compatible enabled advisor, so its quote covers the full selection. For a one-advisor quote, choose a compatible ID from the free advisor list and append ?advisors=id. An x402-capable HTTP client can authorize the exact quoted USDC amount and retry the same panel URL.

Copyable free requests: these commands do not authorize payment or run paid advisors. The trial may run or reuse our own token analysis; the panel request only asks for a quote.

curl -i https://getrugshield.com/v1/advisors
curl -i https://getrugshield.com/v1/trial/token-risk/So11111111111111111111111111111111111111112
curl -i https://getrugshield.com/v1/risk-panel/So11111111111111111111111111111111111111112

For a purchase, supply your actual mint or unsigned transaction, request a fresh unpaid quote for that exact method, URL, query and body, and review its accepts entries. Use a funded wallet client that supports one offered rail; Base payment does not enable Base token analysis. Bound spending to your reviewed amount. Never pay for a sample input merely because it appears here.

Follow the Base and Solana payment-client walkthrough to review a quote, authorize bounded spending and recover the same purchase. Stop for manual review if settlement is uncertain; no automatic review endpoint or replacement payment is prescribed.

Before the first signed request, persist the exact request, original signed proof and a random recovery key. An optional Idempotency-Key is 32 cryptographically random bytes encoded as canonical unpadded base64url (43 characters); use the same key on the first signed request and every recovery retry. The key is optional on the preceding unpaid quote; the client guide creates it after quote review. Keep the returned job ID private. A recovery client must preserve the original proof, rather than generate another authorization.

To customize the panel, read GET /v1/advisors for free and pass selected IDs with ?advisors=id,id. supported_chains names the token networks each advisor analyzes; pays_on names its payment network. The panel currently accepts Solana token mints only. The chain query parameter uses a CAIP-2 network ID (a standard chain identifier) and defaults to solana:5eykt4UsFv8P8NJdTREpY1vzqKqZKvdp. For example, /v1/risk-panel/{mint}?chain=solana:5eykt4UsFv8P8NJdTREpY1vzqKqZKvdp. A Robinhood Chain token such as JEV (eip155:4663) is unsupported; invalid networks and advisor selections return HTTP 400 before a payment quote.

Every discovery surface is public and free to read. There is no signup step or API key to request.

The first signed panel request prepays the exact quote and returns HTTP 202 with charged:true, a confirmed receipt, job_id, status_url, and retry_after_s. Advisors start only after that receipt is saved. Poll the status URL for free until the result is ready; no second signed request or payment is needed. ?prefer_async=true remains accepted for compatibility. A replay returns the original acceptance; polling returns the final result and receipt. The polling interval is not a completion deadline. Paid analysis failures need support review; no automatic partial settlement or refund is promised.

For token and transaction checks, a usable result can settle inline with HTTP 200 and a receipt. An HTTP 503 recovery response can instead report charged:false: poll the returned status URL for free, then repeat the exact original signed request with its recovery key or X-RugShield-Job-ID when instructed. Polling alone never settles this authorization. Do not authorize another payment.

MCP tools are available to payment-aware MCP clients. The first signed token or transaction call can return an answer and receipt without an extra handshake. If a job is still processing, poll get_job_result for free and reuse the original proof with the job ID when ready. The free quote_solana_risk_panel tool returns an estimate and HTTP URL. The HTTP risk panel is not an MCP tool.

Read the next action before sending it.

Fresh JSON guidance reports agent_guidance.payment.status and charge_outcome separately from agent_guidance.analysis.status and result_available. A confirmed charge can coexist with unfinished or failed analysis. Legacy charged:null means unconfirmed, never false; legacy_charged_note explains it. There is no background reconciliation worker (background_reconciliation_available:false).

agent_guidance.request_example describes the recommended next action with type, executable, explanation, effects.payment, effects.analysis, payment_proof_required and new_authorization_allowed. Read both effects before acting. A concrete free request can be executable; templates and original-request references require real inputs or saved context. Manual review and terminal results prescribe no automatic request. An example never authorizes spending for you.

For unknown or uncertain payment outcomes, keep the original references and request support review at rugshield@fastsoftware.uk; do not repay or send keys/proofs by email. Saved replies and receipts are preserved exactly; an older replay may predate these fields. Free job polling supplies current guidance. NO_FLAGS and ALLOW are advisory signals, not proof of safety; no automatic refund is promised.

MCP: https://getrugshield.com/mcp/

• OpenAPI — methods, inputs, outputs and payment metadata
• llms.txt — route selection and payment instructions
• APIs.json — directory-ready service metadata
• x402 discovery — paid resource URLs
• Health — current service status