124 Services from phion.systems

Inspects an MCP tool manifest for security and quality issues including tool name collisions, missing schemas, and prompt injection attacks before installation or trust

Screens a declared subject against official sanctions lists and returns independently retrieved, traceable evidence of matches or clearance

Fetches web content as cryptographically hashed and signed evidence, screened against prompt injection, for use by AI agents

Determines whether an agent action should be allowed to run immediately, held for human approval, or denied outright based on policy rules

Issues a signed, hash-linked receipt attesting to the integrity, provenance, and freshness of AI-agent JSON payloads

Runs a deterministic payment-policy check before an agent executes a payment, returning a signed, short-lived decision receipt authorizing or blocking the transaction.

Validates HTTPS callback and redirect destination URLs against an explicit host allowlist to prevent open redirect and SSRF attacks

Runs independent due diligence on an AI-agent service endpoint, checking x402, A2A, MCP, OpenAPI, ERC-8004, reachability, and returning signed evidence

Verifies a hash-linked agent commerce journey and issues a signed portable receipt proving the chain of events occurred as claimed

Validates least-privilege capabilities, data scope, destinations, budget, and expiry constraints before one agent delegates to another agent

Screens proposed persistent agent-memory writes for poisoning attempts, missing provenance, secret leakage, and unsafe instructions before committing them

Detects and reports changes between a trusted baseline snapshot of a tool's capabilities and its current live manifest

Binds a time-limited human approval decision to a specific purpose and exact payload, creating a cryptographically anchored authorization record for agent actions

Enforce declared per-call, task, session, and period spending budgets before an agent executes a paid action

Validates and cryptographically signs a portable task handoff packet including scope, budget, and expiry for secure agent-to-agent delegation

Resolves a business contact from strong identifiers via People Data Labs and returns signed, verifiable evidence of the enriched data

Makes a deterministic policy decision on whether to proceed with a counterparty transaction based on supplied reputation evidence and transaction limits

Enforces delegated credential scope, requester identity, and expiry constraints without ever receiving or exposing the underlying raw credentials

Checks declared software dependencies for identity integrity, source authenticity, and cryptographic digest coverage

Normalizes public text, HTML, JSON, or XML documents into source-backed, verifiable JSON with provenance tracking

Attests DNS and HTTPS domain-control proof status, producing signed evidence that a declared domain is under the caller's control

Detects compatibility or conflicts between two or more agent policies to prevent unsafe or contradictory agent-to-agent interactions

Identifies changed fields and hashes between two versions of an agent or software manifest

Captures public blockchain explorer or RPC data as immutable evidence with cryptographic content hashes

Rejects impossible or unsafe x402 payment routes, ranks eligible routes by policy, and returns a directly executable next payment action

Resolves a person from strong identifiers via People Data Labs and returns signed, limitation-aware evidence about them

Provides evidence-backed due diligence coverage for real-world asset (RWA) issuers, smart contracts, custody arrangements, and legal documents

Compares a real-world asset's declared NAV, reserves, and collateral against independently retrieved data sources to surface discrepancies

Detects and redacts common secret indicators (API keys, tokens, passwords, credentials) from a payload before it is transmitted to an external tool or service.

Validates a proposed tool call against declared intent, allowed tools, permitted destinations, and cost limits before allowing execution

Inspects and validates untrusted tool output before it enters an agent context or triggers a downstream action

Classifies a failed agent transaction and produces a signed, machine-readable recovery claim with a deterministic remedy plan

Fetches bounded public web content and returns it with source attribution, timestamp, and cryptographic hashes for verifiable provenance

Assesses whether an AI agent tool or external service is still reachable and whether its observations are fresh or stale

Resolves a company via People Data Labs and returns cryptographically signed, limitation-aware enrichment evidence for use by AI agents

Inspects outbound agent payloads for secrets, unsafe destinations, and size limits before they are sent to an external tool or API

Enforces delegated spending mandates by failing closed on inactive mandates and blocking per-transaction or remaining-budget violations

Synthesizes evidence-grounded research on a topic and returns cited findings with signed, verifiable delivery

Detects material state changes in real-world assets including NAV shifts, income distributions, maturity events, redemptions, and fund freezes

Searches bounded public news sources for literal query evidence and returns timestamped, provenance-tagged results

Validates the signature status, freshness, and payload evidence of incoming webhooks to confirm authenticity and integrity

Compares claimed agent or service capabilities against independently evidenced, observed successful outcomes to verify truthfulness

Selects among conflicting agent results using explicit, policy-driven arbitration to resolve ambiguity

Calculate data age against an explicit freshness policy and return a certificate indicating whether the data meets freshness requirements

Detects duplicate charge fingerprints to prevent double-billing in agent payment flows

Calculates a fail-closed maximum cost bound across an agent execution plan to prevent overspending

Detects whether reuse of an idempotency key is a safe replay of a prior request or a conflicting misuse that should be rejected

Verifies a JWT or OAuth token's declared audience (aud) and issuer (iss) claims after local cryptographic signature validation has already been performed

Enriches an entity record with provider-backed data, returning explicit match confidence and enrichment limitations

Commits request, authority, execution, payment, and delivery evidence together into a single auditable proof record

Binds declared user or agent consent to a specific action purpose, scope, and expiry window, creating a verifiable consent record

Returns a deterministic eligibility and transfer-policy decision for real-world asset (RWA) compliance checks

Compares multiple agent results and produces a cryptographically signed record of exact agreement or conflict evidence between them

Search a set of user-supplied public sources and return only results that are literally attested by those sources, with no hallucinated or inferred content

Assesses multiple manipulation signals to produce a sybil/bot reputation score with uncertainty quantification and mandatory abstention when confidence is insufficient

Verifies that an agent holds an active, valid lease on a task before allowing execution or mutation to proceed

Independently verifies authorization, settlement, delivery, and deterministic acceptance of a transaction, then returns a signed portable assurance receipt

Assembles verified failure evidence and recommends dispute review for a transaction or agent interaction without executing any refund or financial action

Recommends an eligible provider under a procurement mandate without executing any purchase

Compares required agent capabilities against offered capabilities before task execution to determine compatibility

Hashes and labels agent context with integrity, confidentiality, and source declarations to establish trust provenance

Hashes and cryptographically attests supplied delivery evidence for tamper-proof verification

Validates agent task envelopes for durability, isolation, and replay-safety before execution

Detects self-referential and reciprocal transaction patterns in economic activity without falsely flagging correlated behavior as fraudulent

Enriches entity fields from bounded, verifiable sources without fabricating missing information

Tests whether source observations meet an explicit maximum-age policy, validating that data is fresh enough to be trusted

Atomically reserves cumulative mandate capacity and returns a signed state receipt for AI-agent payment mandate workflows

Recommends evidence-bound hold or release states for escrow scenarios without taking custody of or moving funds

Bundles observations about a fact from multiple independent sources, exposing where sources agree and disagree

Provider-backed basic entity enrichment with explicit match confidence, limitations, and a signed receipt, paid per-call via x402.

Returns a canonical snapshot or digest-based delta of the PHION tool catalog, including exact execution templates and signed evidence for each listed capability

Executes a bounded, low-cost LLM inference call and returns the result with a cryptographically signed delivery receipt

Validates and cryptographically signs a portable event hash chain to establish tamper-evident audit trails for agent activity

Estimates success probability, quality, latency, and cost for an AI agent service provider, including sample-aware uncertainty quantification

Recommends an auditable, bounded reputation change for an agent or entity based on diverse verified outcome evidence

Verifies asset ownership, payment validity, delivery conditions, and transfer restrictions as a single composite transaction check for real-world asset (RWA) operations

Selects the highest-priority healthy and compatible provider for failover without silent fallback, surfacing the decision transparently

Calculates and cryptographically signs availability and latency evidence derived from bounded samples to produce verifiable SLA attestations

Verifies the cancellation state of an agent task and identifies any unresolved side effects left behind

Binds task state, sequence number, and checkpoint hash into a signed cryptographic evidence record for AI agent task integrity

Validates AI tool output JSON against a bounded top-level JSON Schema subset to ensure structural correctness and safety

Detects material drift between comparable trust vectors without asserting misconduct or making definitive fraud claims

Builds a privacy-bounded trust graph from supplied agent nodes and verified economic relationship edges

Binds a memory record to its source identity, observation timestamp, and content hash to establish verifiable provenance

Negotiates deterministic sustained and burst request-per-second limits for AI agent API access

Returns an evidence-bounded reputation assessment for an AI agent, including confidence score, coverage metrics, and explicit limitations on what the evidence supports

Verifies hash-linked session identity and monotonic sequence continuity for AI agent sessions

Proposes capped shadow prices for services without modifying live production prices

Aggregates diverse verified observations to produce an ERC-8004-compliant reputation score, automatically abstaining when sample size is too low to draw reliable conclusions.

Plans bounded, idempotent resumption of an agent task from a checkpoint after interruption

Captures a timestamped, provenance-tagged public market-data snapshot with freshness metadata

Maps MCP and A2A task identities across protocol boundaries while preserving state, authority, and expiry information

Creates signed cryptographic evidence binding an MCP server's declared domain, endpoint, manifest, public key, and version together into a verifiable identity attestation

Binds an MCP tool-call identity to a payment transaction without silently granting spending authority

Binds verified OAuth issuer and audience claims to an intended receiver, producing signed evidence of the binding

Ranks only tested payment routes based on explicit cost and latency constraints to identify the optimal path

Execute any selected PHION service with a budget cap, persistent idempotency, acceptance checks, and a signed completion envelope

Performs a web search with source attribution, returning normalized results with signed delivery

Detects and stops repeated execution cycles and step count overruns in AI agent workflows

Signs and records a declared data retention or deletion action to produce a verifiable compliance receipt

Normalizes an agent payload to a PHION service schema without altering payment-critical values, returning a signed repair receipt

Estimates the probability of SLA failure based on bounded outcome samples and latency measurements

Preserves attributable public social-source observations with explicit identity limitations as signed evidence

Enforces recurring period budgets for agent subscriptions, blocking spend that exceeds defined limits within a billing cycle

Recommends bounded recovery actions for ambiguous or failed payments before a retry is attempted, helping agents reconcile uncertain transaction states safely.

Estimates evidence-supported probability and expected loss for a transaction to support accept/reject decisions

Normalizes and compares multiple x402 payment quotes to help agents select the best route or validate consistency across providers

Enforces concurrent execution slot limits to prevent agents or workflows from exceeding defined parallelism thresholds

Validates and bundles payment receipt evidence from multiple agents into a verifiable, tamper-evident package without asserting external truth claims

Reconciles payment and receipt fields to verify that payment records match corresponding receipt data

Validates that an x402 payment quote will remain valid long enough to survive the authorization and retry cycle before accepting it

Creates a privacy-minimized behavioral commitment (fingerprint) from bounded agent observations for trust and accountability purposes

Calculates reproducible capability performance scores from versioned, evidence-bound test cases for AI agent services

Bundles protocol receipt commitments across multiple protocols into a single auditable, consistent receipt artifact

Binds an ERC-8004 registry identity to independently marked evidence without asserting ownership claims

Forecasts market demand using verified demand windows, with automatic abstention when sample sizes are too low to produce reliable predictions

Detects inconsistent payment-delivery states, specifically paid-without-delivery and delivered-without-payment conditions, to surface atomicity violations in agentic payment flows

Reports price bands derived exclusively from settled, confirmed market observations

Identifies capability shortages by cross-referencing verified demand signals against available provider supply

Selects a compatible x402 v2 payment requirement for an agent without exposing or transmitting private keys

Binds an A2A (Agent-to-Agent) task identity to a financial transaction while preserving explicit authority gaps between agents

Validates whether an MCP server implementation is compatible with the 2026 MCP protocol specification, including method and stateless-discovery requirements

Binds an MCP catalog digest to a TTL, scope, and revalidation policy to control caching behavior for AI-agent tool catalogs

Maps and reviews mandate fields for AP2-style authorization workflows without claiming full AP2 production control authority