Correctness-critical services for autonomous agents

Returns a 0-100 reliability score for another x402 service (availability, latency, fidelity, and identity sub-scores) computed from an append-only, hash-anchored probe evidence trail. Use it to vet a counterparty x402 endpoint's trustworthiness before paying it.

Returns the reliability-score time series for an x402 service (most recent first), so an agent can judge trend and stability rather than a single point-in-time score.

Extracts structured JSON from unstructured text/markdown/HTML according to a caller-supplied JSON Schema. Cost-guarded so the price always exceeds the model token cost; oversized input and plaintext PII/secrets are rejected before settlement (never billed, never forwarded to the model).

Fetches an owned, continuously-maintained niche data feed. Current feed `x402-directory`: a probed directory of known x402 services with their live reliability scores, networks, and last-probe timestamps.

Redacts PII and secrets (emails, phone numbers, SSNs, credit cards, API keys, JWTs, private keys) from caller content and returns the redacted text plus an off-chain Ed25519-signed attestation binding input to output. Choose how spans are replaced: `remove` (default `[REDACTED]`), `mask` (format-preserving partial masking), or `pseudonymize` (stable keyed HMAC tokens). The attestation is independently verifiable at /v1/attest/{id}; no content is forwarded to any model.

Returns a previously issued redaction attestation (public key, signature, content hash, detector version) so any party can independently verify a /v1/redact result. Free.

Returns the current Ed25519 public key (SPKI PEM) used to sign /v1/redact attestations, so any party can verify a redaction proof offline without first holding an attestation id. Free.

Vets an external x402 service before you pay it: fetches the target's /.well-known/x402 manifest (who is paid, which chain/scheme), runs a live health probe (reachability + latency), and returns our own append-only reliability score for that host if we have probed it. One paid call to decide whether a counterparty endpoint is safe to settle against.

Validates a JSON value against a caller-supplied JSON Schema and returns a definitive valid/invalid verdict with precise, JSON-Pointer-addressed errors. Use it to gate data handed between tools or returned by an LLM before acting on it. Pure, deterministic, no model involved.

Funds prepaid State Hub credit with a single x402 settlement and returns a wallet-scoped tenant bearer key. One deposit then covers many micro-priced ops (save/query/search/cache) billed in-DB against the credit — pay settlement overhead once, then run durable agent memory cheaply at scale. Top up any time by depositing again.

Returns the remaining prepaid State Hub credit (USD) for the wallet that owns the supplied tenant key. Free (no x402 charge); authenticate with the tenant bearer key.

Stores a JSON document under a key, scoped to the paying wallet, with a TTL. Priced by TTL tier and spent against prepaid credit (fund once via POST /v1/state/deposit). Every doc is screened at write time (PII/secrets/opaque blobs) per GET /v1/policy. Tier-0 inline screen returns in <50 ms; deeper review is async.

Reads wallet-scoped state by exact key or key prefix. Spent against prepaid credit (fund once via POST /v1/state/deposit). Returns only non-expired rows belonging to the authenticated wallet (RLS-isolated).

Deletes one state key for the authenticated wallet. Free (no charge); authenticate with the tenant bearer key.

Redis hot cache set with TTL (wallet-scoped). Spent against prepaid credit. Returns 503 when Redis is unavailable — Postgres state endpoints still work.

Semantic nearest-neighbor search over wallet-scoped vectors (pgvector, RLS-isolated). Opt in to embedding on save with embed:true. Premium prepaid-credit tier.