{"name":"bluefox-edge","version":"2.0.0","description":"Brand claim verification and monitoring for AI agents","url":"/mcp","auth":{"scheme":"bearer","header":"Authorization: Bearer <key>","help_url":"https://api.bluefoxedge.ai/v1/auth/help"},"payment":{"protocol":"x402","networks":["eip155:8453"],"currency":"USDC","note":"Priced REST verification endpoints accept x402 payment (USDC over the x402 protocol; each priced endpoint's 402 response states its exact terms) as an alternative to an API key; the interactive /mcp endpoint and the /mcp/tools/* shims require an API key (Authorization: Bearer <key>)"},"rest_only_endpoints":[{"endpoint":"POST /v1/notarize/hash","description":"Submit a sha256 hex digest; receive a signed receipt recording that the digest was presented at that time (content never seen or judged). Fully keyless: payable per request via x402, no account needed. REST-only — not an MCP tool; see /openapi.json and /.well-known/x402."}],"tools":[{"name":"bluefox_check_brand","category":"verification","description":"Verify a specific factual claim about a brand's identity, ownership, positioning, or market presence against BlueFox Edge's reference database. Use this when you need to fact-check any brand-related statement before presenting it to a user. Input: brand slug and one or more claims as strings. Returns: verdict (accurate, inaccurate, partially_accurate, unverifiable), confidence score (0-1), and evidence text for each claim. Covers 470+ brands with curated definitions — distinct from bluefox_check_price (pricing), bluefox_check_policy (policies), and bluefox_check_financial (financials) which handle domain-specific claims. Returns verified result in ~150 tokens (vs ~4,500 tokens for web search equivalent).","parameters":{"brand":"(required, string) Brand slug, lowercase hyphenated: 'nike', 'coca-cola', 'best-buy'. Use bluefox_get_brand to look up available slugs.","claims":"(required, array of strings) One or more factual claims to verify. Each claim is verified independently. Example: ['Nike offers free returns', 'Nike was founded in 1964']","depth":"(optional, string) Verification depth. 'fast' = rule engine only (<15ms). 'standard' = LLM verification (~525ms). 'auto' = tries rules first, escalates to LLM if uncertain. Default: 'auto'"}},{"name":"bluefox_check_price","category":"verification","description":"Verify whether a product pricing claim is accurate by comparing it against BlueFox Edge's tracked price database. Use this when an agent states a price, discount, or cost for any product. Input: brand slug and one or more pricing claims as strings. Returns: verdict, confidence score, the tracked price (historical — as of 2026-06-11), a staleness flag indicating data freshness, and the data source. Tracks 2,300+ products across 437 brands — distinct from bluefox_check_brand which handles identity claims, not pricing. Returns verified result in ~150 tokens (vs ~4,500 tokens for web search equivalent).","parameters":{"brand":"(required, string) Brand slug, lowercase hyphenated: 'apple', 'nike', 'samsung'.","claims":"(required, array of strings) Pricing claims to verify. Include the product name and price. Example: ['MacBook Air M4 costs $1,099', 'AirPods Pro cost $249']","depth":"(optional, string) 'fast', 'standard', or 'auto'. Default: 'auto'"}},{"name":"bluefox_check_policy","category":"verification","description":"Verify whether a claim about a brand's return, shipping, warranty, or cancellation policy is accurate. Use this when an agent states a return window, shipping cost, warranty term, or other policy detail. Input: brand slug and one or more policy claims as strings. Returns: verdict, confidence score, and the actual policy data from BlueFox Edge's structured records. Covers 465+ brands — distinct from bluefox_check_price (pricing) and bluefox_check_compliance (regulatory status). Returns verified result in ~150 tokens (vs ~4,500 tokens for web search equivalent).","parameters":{"brand":"(required, string) Brand slug, lowercase hyphenated.","claims":"(required, array of strings) Policy claims to verify. Example: ['Nike offers 60-day returns', 'Free shipping on orders over $50']","depth":"(optional, string) 'fast', 'standard', or 'auto'. Default: 'auto'"}},{"name":"bluefox_check_financial","category":"verification","description":"Verify whether a financial claim about a company is accurate by checking against SEC filings and curated financial data. Use this when an agent states revenue, net income, market cap, CEO, headquarters, or founding year for a publicly traded brand. Input: brand slug and one or more financial claims as strings. Returns: verdict, confidence score, and the actual financial data with source attribution. Covers 287+ brands with data sourced from SEC filings — distinct from bluefox_check_brand which handles identity claims, not financial metrics. Returns verified result in ~150 tokens (vs ~4,500 tokens for web search equivalent).","parameters":{"brand":"(required, string) Brand slug, lowercase hyphenated.","claims":"(required, array of strings) Financial claims to verify. Example: [\"Tesla's revenue grew 20% last year\", 'Tim Cook is CEO of Apple']","depth":"(optional, string) 'fast', 'standard', or 'auto'. Default: 'auto'"}},{"name":"bluefox_check_spec","category":"verification","description":"Verify whether a product specification claim is accurate by checking against extracted spec sheets. Use this when an agent states a technical spec like RAM, storage, weight, dimensions, or battery life for a product. Input: brand slug and a single specification claim as a string. Returns: verdict, confidence score, the actual spec value, and known product variants. Covers 660+ products — distinct from bluefox_check_price (pricing) and bluefox_check_brand (identity). Returns verified result in ~150 tokens (vs ~4,500 tokens for web search equivalent).","parameters":{"brand":"(required, string) Brand slug, lowercase hyphenated.","claim":"(required, string) A single product specification claim. Include the product name and the spec value. Example: 'MacBook Pro M4 has 36GB unified memory'","depth":"(optional, string) 'fast', 'standard', or 'auto'. Default: 'auto'"}},{"name":"bluefox_check_compliance","category":"verification","description":"Verify whether a regulatory or compliance claim about a brand is accurate, detecting common misrepresentations like 'FDA approved' for supplements. Use this when an agent makes statements about a brand's FDA, GDPR, CCPA, HIPAA, SOC 2, or ISO 27001 status. Input: brand slug and one or more compliance claims as strings. Returns: verdict, confidence score, and the actual compliance status with regulatory context. Covers 420+ brands — distinct from bluefox_check_policy (returns/shipping) and bluefox_check_brand (identity). Returns verified result in ~150 tokens (vs ~4,500 tokens for web search equivalent).","parameters":{"brand":"(required, string) Brand slug, lowercase hyphenated.","claims":"(required, array of strings) Compliance claims to verify. Example: ['Salesforce is SOC 2 certified', 'Apple is GDPR compliant']","depth":"(optional, string) 'fast', 'standard', or 'auto'. Default: 'auto'"}},{"name":"bluefox_check_competitive","category":"verification","description":"Verify a comparative claim between two or more brands using cross-model AI consensus data. Use this when an agent compares brands or makes relative statements like 'X is more popular than Y' or 'X outsells Y.' Input: one or more comparative claim strings naming at least two brands. Returns: verdict, confidence score, and evidence from AI model consensus. Based on data from 156+ AI models — distinct from bluefox_check_brand which checks single-brand facts against reference data. Returns verified result in ~150 tokens (vs ~4,500 tokens for web search equivalent).","parameters":{"claims":"(required, array of strings) Comparative claims between brands. The claim should name at least two brands. Example: ['Tesla outsells BMW in the US', 'Netflix has more subscribers than Disney+']","depth":"(optional, string) 'fast', 'standard', or 'auto'. Default: 'auto'"}},{"name":"bluefox_check_sentiment","category":"verification","description":"Analyze the sentiment of one or more text strings to determine if they are positive, negative, or neutral. Use this when you need to assess the tone of user reviews, agent responses, or brand mentions before acting on them. Input: an array of text strings to analyze. Returns: sentiment label (positive, negative, or neutral) and confidence score (0-1) for each input string. Runs on a BERT-family model with ~50ms latency per call. Returns verified result in ~150 tokens (vs ~4,500 tokens for web search equivalent).","parameters":{"texts":"(required, array of strings) One or more text strings to analyze for sentiment. Each string is scored independently."}},{"name":"bluefox_check_freshness","category":"verification","description":"Check whether sources, citations, and date references in text are current or stale. Use this before presenting any agent response that cites specific dates, statistics, or named sources to catch outdated information. Input: a text string containing citations or date references. Returns: freshness verdict for each detected reference, staleness flags, and suggested updates when BlueFox Edge has newer data. Supports fast mode (regex, <20ms) and standard mode (LLM cross-reference, ~1s). Returns verified result in ~150 tokens (vs ~4,500 tokens for web search equivalent).","parameters":{"text":"(required, string) Response text containing citations, date references, or named sources to check for freshness.","depth":"(optional, string) 'fast' = regex pattern matching (<20ms). 'standard' = LLM analysis with cross-reference (~1s). 'auto' = fast first, LLM if needed. Default: 'auto'"}},{"name":"bluefox_check_customs","category":"verification","description":"Check whether multiple independent AI models AGREE on the HS/HTS tariff classification of a shipment or product description. Use this when an agent needs a cross-model consistency signal on a customs code before quoting duties or preparing a filing. Input: one free-text shipment/product description. Returns: consistency verdict (unanimous, majority, or unresolved), the agreed code set at the declared comparison depth, per-model vote tally with output digests, and the pinned consensus policy + tariff edition identifiers. The verdict is AGREEMENT, never correctness — no accuracy figure exists for this dimension and none is implied. The description is bound by digest and never stored; model output text is never returned or persisted. Every authenticated check mints one signed reliance receipt (the receipt id rides the response). Distinct from bluefox_check_compliance, which verifies regulatory claims — this tool classifies goods.","parameters":{"description":"(required, string) The shipment/product description to classify, 1-10,000 characters."}},{"name":"bluefox_preflight_check","category":"safety","description":"Run a pre-flight safety check on an agent response before delivering it to the user. Use this as the final gate before showing any agent-generated text — it catches safety issues, factual errors, brand misalignment, bias, and PII exposure. Input: the response text, optionally the original user query and brand slug for context. Returns: pass/fail verdict with per-dimension scores across 6 dimensions (safety, accuracy, brand alignment, professionalism, bias, PII). Supports fast mode (rule-based, <20ms) and standard mode (full LLM evaluation, ~1-2s) — distinct from bluefox_check_safety which classifies content against GARM advertising categories. Returns verified result in ~150 tokens (vs ~4,500 tokens for web search equivalent).","parameters":{"response":"(required, string) The full agent response text to evaluate.","query":"(optional, string) The original user query that triggered this response. Improves accuracy of brand alignment scoring.","brand":"(optional, string) Brand slug for context-specific evaluation. If provided, checks brand alignment and messaging consistency.","depth":"(optional, string) 'fast' = rule-based checks only (<20ms). 'standard' = full LLM evaluation with claim verification (~1-2s). Default: 'standard'"}},{"name":"bluefox_preflight_batch","category":"safety","description":"Run pre-flight safety checks on multiple agent responses in a single call for batch processing. Use this when you have 2 or more responses to evaluate simultaneously instead of calling bluefox_preflight_check repeatedly. Input: an array of check objects, each containing a response string and optional query/brand context. Returns: an array of pass/fail results with per-dimension scores, one per input. Same 6-dimension evaluation as bluefox_preflight_check — use the single-check variant for one-off evaluations. Returns verified result in ~150 tokens (vs ~4,500 tokens for web search equivalent).","parameters":{"checks":"(required, array of objects) Each object has: 'response' (required, string), 'query' (optional, string), 'brand' (optional, string).","depth":"(optional, string) 'fast' or 'standard'. Default: 'standard'"}},{"name":"bluefox_check_safety","category":"safety","description":"Classify content against the 12 GARM (Global Alliance for Responsible Media) brand safety categories for advertising suitability. Use this when you need to determine whether content is safe for brand-adjacent advertising, sponsorship, or media placement. Input: the content text, optionally a brand slug and content type (article, comment, ad, text). Returns: per-category risk scores, floor violations, and an overall safety verdict. Covers all 12 GARM categories — distinct from bluefox_preflight_check which evaluates agent response quality, not advertising suitability. Returns verified result in ~150 tokens (vs ~4,500 tokens for web search equivalent).","parameters":{"content":"(required, string) Text content to classify for brand safety.","brand":"(optional, string) Brand slug for context-specific scoring. Some brands have stricter safety thresholds.","content_type":"(optional, string) Type of content: 'article', 'comment', 'ad', 'text'. Default: 'text'","depth":"(optional, string) 'fast' = keyword scan (<10ms). 'standard' = LLM classification (~1s). 'auto' = keyword first, LLM if uncertain. Default: 'auto'"}},{"name":"bluefox_egress_gate","category":"safety","description":"Run the fail-closed egress gate over an outbound payload BEFORE it leaves your fleet. The pinned, versioned PHI detector (phi_scan) scans the content across its declared detection classes (a declared subset of the 18 HIPAA identifiers — coverage map, limits, and a measured accuracy row ride every response): decision 'proceed' iff nothing flagged, else 'block' with per-class reasons; a detector that cannot run is a 503 halt, never a silent pass. Every check mints one signed reliance receipt binding the payload by sha256 digest (the content itself is never stored) — the decision gates in-line, the signed receipt materializes seconds later. Voluntary check surface: receipts prove the calls that happened. Distinct from bluefox_preflight_check (composite response quality) and bluefox_check_safety (GARM advertising categories).","parameters":{"content":"(required, string) The egress payload text to check, 1-10,000 characters."}},{"name":"bluefox_get_brand","category":"brand-data","description":"Retrieve the raw brand definition record for a given brand slug, including positioning, policies, pricing rules, and key messaging. Use this when you need the underlying data BlueFox Edge holds for a brand, for programmatic access or debugging. Input: brand slug as a string. Returns: the full brand definition object with all fields (name, category, positioning, policies, pricing, messaging). Covers 470+ brands — for agent-friendly structured profiles, use bluefox_get_brand_profile instead. Returns verified result in ~150 tokens (vs ~6,000 tokens for web search equivalent).","parameters":{"slug":"(required, string) Brand slug, lowercase hyphenated. Examples: 'nike', 'coca-cola', 'best-buy'. Use the /brands endpoint to list all available slugs."}},{"name":"bluefox_get_brand_profile","category":"brand-data","description":"Get a comprehensive, agent-ready brand profile with structured sections for easy consumption. Use this when an agent needs full context about a brand before answering questions — it provides everything in one call. Input: brand slug as a string. Returns: structured profile with identity, policies (returns/shipping/warranty), pricing, financials, contact info, business hours, availability, promotions, and a completeness score. Cached for 1 hour — distinct from bluefox_get_brand which returns the raw definition record. Returns verified result in ~150 tokens (vs ~6,000 tokens for web search equivalent).","parameters":{"slug":"(required, string) Brand slug, lowercase hyphenated. Examples: 'nike', 'apple', 'starbucks'."}},{"name":"bluefox_register_agent","category":"utility","description":"Register a new AI agent with BlueFox Edge to begin building a verification trust score over time. Use this once during agent setup — include the returned token in subsequent verification requests to accumulate trust history. Input: agent display name and optional description of what the agent does. Returns: agent_id and agent_token for authentication. Trust scores are public and based on 5 factors: accuracy, consistency, volume, recency, and brand diversity — this is a one-time setup call, not a workflow tool.","parameters":{"name":"(required, string) Display name for the agent. Example: 'ShopBot Pro'","description":"(optional, string) Brief description of what the agent does. Example: 'E-commerce shopping assistant that recommends products'"}},{"name":"bluefox_brand_health","category":"monitoring","description":"Get a health snapshot for a brand based on how 156+ AI models discuss it. Use this when you need a quick overview of a brand's standing across the AI ecosystem. Input: brand slug as a string. Returns: health score (0-100), sentiment breakdown (positive/neutral/negative), recommendation rate, mention count, model agreement rate, trends, and active alerts. Computed daily from 2M+ AI model responses — for deeper analysis, use bluefox_brand_narratives, bluefox_brand_models, or bluefox_brand_insights. Returns verified result in ~150 tokens (vs ~6,000 tokens for web search equivalent).","parameters":{"slug":"(required, string) Brand slug, lowercase hyphenated. Examples: 'nike', 'tesla', 'starbucks'. Use /brands to list available slugs."}},{"name":"bluefox_brand_narratives","category":"monitoring","description":"Get the top claims and themes that AI models associate with a brand, ranked by frequency. Use this when you need to understand what AI models are actually saying about a brand and which narratives are gaining or losing traction. Input: brand slug and optional limit (1-100, default 20). Returns: ranked list of narratives, each with claim text, frequency count, sentiment, trend direction (rising/falling/stable), and an emerging flag. Distinct from bluefox_brand_health which gives aggregate scores — this shows the specific claims driving those scores. Returns verified result in ~150 tokens (vs ~6,000 tokens for web search equivalent).","parameters":{"slug":"(required, string) Brand slug, lowercase hyphenated.","limit":"(optional, integer) Maximum number of narratives to return. Range: 1-100. Default: 20"}},{"name":"bluefox_brand_models","category":"monitoring","description":"Get per-model sentiment breakdown showing how each AI model (GPT-4, Claude, Gemini, Grok, etc.) views a brand individually. Use this to detect when a specific AI model has an unusually positive or negative view compared to the cross-model mean. Input: brand slug as a string. Returns: per-model sentiment scores, recommendation rates, mention counts, divergence flags (>0.2 std dev), and the cross-model mean. Distinct from bluefox_brand_health which aggregates across models — this reveals which specific models are outliers. Returns verified result in ~150 tokens (vs ~6,000 tokens for web search equivalent).","parameters":{"slug":"(required, string) Brand slug, lowercase hyphenated."}},{"name":"bluefox_brand_recommendations","category":"monitoring","description":"Get how frequently AI models recommend a brand across different product categories. Use this to understand a brand's competitive position in AI-generated recommendations and track changes over time. Input: brand slug as a string. Returns: per-category recommendation rates, 30-day trend direction, and the top competitor in each category. Distinct from bluefox_competitive_radar which compares health scores head-to-head — this focuses specifically on recommendation frequency by category. Returns verified result in ~150 tokens (vs ~6,000 tokens for web search equivalent).","parameters":{"slug":"(required, string) Brand slug, lowercase hyphenated."}},{"name":"bluefox_competitive_radar","category":"monitoring","description":"Compare a brand's health score and recommendation rate against specified competitors in a side-by-side view. Use this when you need to benchmark a brand against its competitors across AI model perception. Input: primary brand slug and comma-separated competitor slugs (max 10). Returns: health score and recommendation rate for the primary brand and each competitor. Distinct from bluefox_brand_recommendations which shows one brand's category-level detail — this provides head-to-head comparison across brands. Returns verified result in ~150 tokens (vs ~4,500 tokens for web search equivalent).","parameters":{"slug":"(required, string) Primary brand slug to analyze.","competitors":"(required, string) Comma-separated competitor slugs. Example: 'adidas,puma,new-balance'. Maximum 10 competitors."}},{"name":"bluefox_brand_insights","category":"monitoring","description":"Get AI-generated actionable insights and strategic recommendations for a brand based on its monitoring data. Use this when you want a synthesized analysis rather than raw metrics — it interprets health scores, sentiment trends, narratives, and competitive position. Input: brand slug as a string. Returns: a list of prioritized insights with recommended actions. Cached for 24 hours — distinct from bluefox_brand_health which returns raw metrics, this provides interpreted recommendations. Returns verified result in ~150 tokens (vs ~6,000 tokens for web search equivalent).","parameters":{"slug":"(required, string) Brand slug, lowercase hyphenated."}},{"name":"bluefox_track_claim","category":"monitoring","description":"Register one or more claims to monitor for accuracy across AI models over time. Use this to set up ongoing monitoring of specific brand facts — BlueFox Edge will track how often AI models make this claim and whether they get it right. Input: brand slug and an array of claim strings or objects with expected verdicts. Returns: confirmation with tracking IDs for each registered claim. This is a setup action for ongoing monitoring — use bluefox_check_brand for one-time claim verification. Returns verified result in ~150 tokens (vs ~3,500 tokens for web search equivalent).","parameters":{"slug":"(required, string) Brand slug the claims are about.","claims":"(required, array) Claims to track. Each item can be a string or an object with 'claim' (string) and 'expected_verdict' ('accurate', 'inaccurate', etc.)."}},{"name":"bluefox_subscribe_monitor","category":"monitoring","description":"Create a webhook subscription for alerts when a brand's monitoring data changes significantly. Use this to receive automatic notifications when sentiment drops, recommendation rates change, or model divergence is detected. Input: brand slug, an HTTPS webhook URL, and optional trigger conditions. Returns: subscription ID and confirmation. Supports conditions: sentiment_drop, recommendation_drop, model_divergence, any_alert — distinct from bluefox_track_claim which monitors specific claims, this monitors overall brand health changes.","parameters":{"slug":"(required, string) Brand slug to monitor.","webhook_url":"(required, string) HTTPS URL that will receive POST webhook notifications. Must be a public HTTPS URL (no localhost or private IPs).","conditions":"(optional, array of strings) Trigger conditions. Options: 'sentiment_drop', 'recommendation_drop', 'model_divergence', 'any_alert'. Default: ['sentiment_drop', 'recommendation_drop']"}},{"name":"edge_advisor_recommend","category":"advisor","description":"Find the best MCP server for your requirements based on continuous behavioral observation data. Use this when you need to select an MCP server for a task and want recommendations ranked by reliability, security, and capability match. Input: optional category, capabilities list, constraints (transport, auth, max_tools), and ranking preferences. Returns: ranked list of MCP server recommendations with scores across maintenance health, behavioral consistency, and security dimensions. Distinct from edge_advisor_compare which does side-by-side comparison of specific servers you already know about. Returns verified result in ~150 tokens (vs ~5,000 tokens for web search equivalent).","parameters":{"category":"(optional, string) Server category (e.g., 'browser-automation', 'database')","capabilities":"(optional, array of strings) Required capabilities","constraints":"(optional, object) Hard filters: transport, auth, max_tools, self_hostable, min_maintenance_score, max_vulnerability_count, ecosystems","preferences":"(optional, object) Ranking preferences: minimize (scope_surface, dependency_count), maximize (maintenance_health, behavioral_consistency)","detail":"(optional, string) Response detail: 'summary' or 'full' (default: 'full')"}},{"name":"edge_advisor_compare","category":"advisor","description":"Compare 2-5 MCP servers side by side across all observation dimensions including security, maintenance, and behavioral consistency. Use this when you have specific MCP servers in mind and need a detailed comparison to choose between them. Input: comma-separated entity IDs (2-5 servers). Returns: side-by-side metrics for each server across all scored dimensions. Distinct from edge_advisor_recommend which discovers new servers — this compares servers you already know about. Returns verified result in ~150 tokens (vs ~5,000 tokens for web search equivalent).","parameters":{"entities":"(required, string) Comma-separated entity IDs (2-5)"}},{"name":"edge_advisor_categories","category":"advisor","description":"List all available MCP server categories with entity counts and data readiness indicators. Use this to browse what types of MCP servers are catalogued before using edge_advisor_recommend to find specific servers. Input: none required. Returns: list of categories, each with a name, entity count, and data readiness flag. This is a discovery tool — use edge_advisor_recommend after identifying a relevant category. Returns verified result in ~150 tokens (vs ~5,000 tokens for web search equivalent).","parameters":{}},{"name":"edge_ops_latest","category":"ops","description":"Get the latest operational intelligence events from a specific feed. Use this to stay current on CVEs, price changes, compliance updates, protocol changes, or model behavior shifts relevant to AI agents. Input: feed slug (security_ops, commerce_ops, protocol_ops, trust_ops, regulatory_ops, model_ops), optional priority filter, and limit. Returns: list of events with event_id, type, priority, headline, summary, and timestamp. For critical-only alerts across all feeds, use edge_ops_check instead. Returns verified result in ~150 tokens (vs ~2,000 tokens for web search equivalent).","parameters":{"feed":"(optional, string) Feed slug: security_ops, commerce_ops, protocol_ops, trust_ops, regulatory_ops, model_ops. Default: security_ops","priority":"(optional, string) Filter by priority: P1 (critical), P2 (important), P3 (informational)","limit":"(optional, integer) Max events (default: 10, max: 50)"}},{"name":"edge_ops_check","category":"ops","description":"Get all P1 (critical) ops events across all feeds from the last 24 hours in a single zero-config call. Use this as a quick safety check before starting any agent workflow — it surfaces critical threats, outages, and compliance changes without requiring feed selection. Input: none required. Returns: count and list of P1 events from all feeds with event_id, feed, type, headline, summary, and timestamp. This is the free-tier entry point — use edge_ops_latest for filtered access to specific feeds and priority levels. Returns verified result in ~150 tokens (vs ~2,000 tokens for web search equivalent).","parameters":{}},{"name":"edge_ops_subscribe","category":"ops","description":"Subscribe to ops feed webhooks to receive event notifications for operational intelligence events (Pro+ plan required). Use this to set up automated alerting for security vulnerabilities, price changes, or compliance updates relevant to your agent workflows. Input: HTTPS webhook URL, optional comma-separated feed slugs (empty = all feeds), and priority filter. Returns: subscription ID, webhook secret for HMAC-SHA256 signature verification, and confirmation. Distinct from bluefox_subscribe_monitor which monitors brand health — this monitors the operational threat landscape.","parameters":{"url":"(required, string) HTTPS webhook URL to receive POST notifications","feeds":"(optional, string) Comma-separated feed slugs to subscribe to. Empty = all feeds.","priority_filter":"(optional, string) Comma-separated priorities: P1, P2, P3. Default: P1 only"}},{"name":"edge_pre_transaction_verify","category":"verification","description":"Run a comprehensive trust check before completing a purchase or transaction. Verifies the brand, price, policies, and compliance in a single call. Use this when your agent is about to make or recommend a purchase and needs to confirm the transaction is trustworthy. Input: brand name, optional product, price, and specific claims (max 10). Returns: overall trust verdict (proceed/caution/block), per-domain results, and recommendation. Runs up to 6 verification checks; saves ~18,000 tokens vs performing each check manually.","parameters":{"brand":"(required, string) Brand name to verify","product":"(optional, string) Product name for price/spec verification","price":"(optional, object) {amount: number, currency: string} — claimed price to verify","claims":"(optional, array of strings, max 10) Specific claims to verify","checks":"(optional, array of strings, max 6) Verification domains: brand, price, policy, compliance, spec, financial"}},{"name":"edge_verify_protocol_claim","category":"verification","description":"Verify policy claims from commerce protocols (UCP, ACS, ACP) against BlueFox Edge's independent reference data. Submit claims about return policies, shipping policies, or warranties that a merchant has published through a commerce protocol, and Edge will verify accuracy against its verified reference database of 1,000+ brands. Supported claim types in Phase 1: return_policy, shipping_policy, warranty. Returns per-claim match levels (exact/partial/mismatch/unverified/unsupported), discrepancy descriptions, freshness timestamps, and confidence levels.","parameters":{"brand":"(required, string) Brand slug, lowercase hyphenated (e.g. 'nike').","claims":"(required, array of objects, max 10) Each claim: {source_protocol: string (ucp/acp/acs/mcp/other), claim_type: string (return_policy/shipping_policy/warranty), claim_value: string (max 1000 chars), claim_timestamp: optional ISO 8601 string}."}},{"name":"edge_create_watcher","category":"monitoring","description":"Create a custom watcher that monitors a brand for specific changes and delivers structured events when conditions are met. Use when your agent needs proactive notifications about brand changes rather than repeated polling. Input: brand slug, array of conditions (field + trigger + optional threshold/value), optional webhook URL, cooldown period, and priority filter. Returns: watcher ID, webhook secret (if webhook configured), and full watcher configuration. Supports 13 watchable fields (trust_score, return_policy, product_price, etc.) and 7 trigger types (changed, dropped_below, rose_above, equals, not_equals, any, contains). Saves ~4,500 tokens per change detection vs manual polling.","parameters":{"api_key":"(required, string) Your API key (bfx_live_*) — identifies the owning account; watchers are scoped to it.","brand":"(required, string) Brand slug to monitor, e.g. 'nike'","conditions":"(required, array of objects) 1-5 conditions. Each: {field: string, trigger: string, threshold?: number, value?: string}","operator":"(optional, string) 'AND' or 'OR' for combining conditions. Default: 'OR'","webhook_url":"(optional, string) HTTPS URL for push delivery","priority_minimum":"(optional, string) Minimum priority to trigger: 'P1', 'P2', or 'P3'. Default: 'P3'","cooldown_seconds":"(optional, integer) Minimum seconds between firings. Default: 3600"}},{"name":"edge_list_watchers","category":"monitoring","description":"List your custom watchers with optional filters by brand, status, or enabled state. Use to review active surveillance rules or find watchers for a specific brand. Returns a paginated list (ordered newest-first by creation time) with watcher configurations, trigger counts, and last triggered timestamps. Saves ~2,000 tokens vs calling individual watcher endpoints.","parameters":{"api_key":"(required, string) Your API key (bfx_live_*) — only watchers owned by this key are returned.","brand":"(optional, string) Filter by brand slug","status":"(optional, string) Filter: 'active', 'paused', 'expired', 'error'","enabled":"(optional, boolean) Filter by enabled state","limit":"(optional, integer) Results per page, max 100. Default: 20","offset":"(optional, integer) Pagination offset. Default: 0"}},{"name":"edge_get_watcher_events","category":"monitoring","description":"Get triggered events for a specific watcher. Use to poll for changes when webhook delivery is not configured or as a backup to webhooks. Returns structured change events with before/after values, confidence scores, and evidence. Events are marked as retrieved on poll. Saves ~4,500 tokens per event vs manual verification polling.","parameters":{"api_key":"(required, string) Your API key (bfx_live_*) — must own the watcher being polled.","watcher_id":"(required, string) Watcher ID, e.g. 'w_abc123'","since":"(optional, string) ISO datetime to fetch events from. Default: last 24 hours","limit":"(optional, integer) Max events to return, max 100. Default: 20"}},{"name":"edge_brand_analytics_summary","category":"analytics","description":"Get a summary of how AI agents interact with a brand over the last 30 days. Use this to understand query volume, top verification domains, average confidence, and tokens saved for any brand in BlueFox Edge's database. Input: brand slug. Returns: total queries, unique agents, top 5 domains by query count, average confidence score, and total tokens saved. Requires Starter tier or above. Response is ~200 tokens.","parameters":{"brand":"(required, string) Brand slug, lowercase hyphenated: 'nike', 'apple', 'amazon'"}},{"name":"edge_brand_analytics_trends","category":"analytics","description":"Get week-over-week query trends for a brand showing how agent interest is changing over time. Use this to track volume shifts, agent count changes, confidence movements, and emerging verification domains. Input: brand slug and number of weeks (1-12). Returns: weekly data points with volume change percentages, agent count changes, and emerging domains. Requires Pro tier or above. Response is ~300 tokens.","parameters":{"brand":"(required, string) Brand slug, lowercase hyphenated: 'nike', 'apple', 'amazon'","weeks":"(optional, integer) Number of weeks to return. Default: 4, max: 12"}},{"name":"edge_check_reputation","category":"agent","description":"Check your agent's behavioral reputation score on BlueFox Edge. Returns a multi-dimensional reliability assessment based on query patterns, verification usage, rate compliance, and interaction history. Authenticate with your agent token (from POST /agents/register). Use this to understand your agent's standing and identify areas for improvement. Returns: reputation score (0-100), tier (new/emerging/established/trusted/verified), confidence level, five dimension scores, interaction history, and criteria to reach the next tier. Response is ~300 tokens.","parameters":{"agent_token":"(required, string) Your agent token (bfa_live_*) from POST /agents/register."}}]}