MCP Tools
What is MCP?
The Model Context Protocol (MCP) is an open standard for connecting AI models to external tools and data sources. Instead of calling a REST API directly, your agent discovers and calls tools through MCP — the client handles transport, authentication, and response formatting.
Connect
Edge serves its tools over Remote MCP (Streamable HTTP) at a single endpoint. Add it to Claude Desktop, Cursor, VS Code, or any MCP client with this config:
{
"mcpServers": {
"bluefox-edge": {
"url": "https://api.bluefoxedge.ai/mcp",
"headers": {
"Authorization": "Bearer YOUR_API_KEY"
}
}
}
}Use the endpoint exactly as shown — https://api.bluefoxedge.ai/mcp with no trailing slash. No local process or npm package is required; the remote server handles tool execution.
Need a key? Get a free API key by signing up with your email — no credit card required. The free tier includes 100 API calls per UTC day overall — the one quota that binds MCP tool calls. (The 10 claim-verification calls/day inner meter applies only to the REST /verify endpoints, never to MCP tool calls.)
All 41 Tools
- bluefox_check_brand— Verify a factual claim about a brand's identity, ownership, or positioning
- bluefox_check_price— Verify a product pricing claim
- bluefox_check_policy— Verify a return, shipping, warranty, or store-policy claim
- bluefox_check_financial— Verify a financial claim about a company
- bluefox_check_spec— Verify a product specification claim
- bluefox_check_compliance— Verify a regulatory or compliance claim
- bluefox_check_competitive— Verify a comparative claim between two or more brands
- bluefox_check_sentiment— Analyze the sentiment of one or more text strings
- bluefox_check_freshness— Check whether sources, citations, and dates in text are current
- bluefox_check_customs— Cross-model consistency check on the HS/HTS tariff classification of a product
- edge_verify_protocol_claim— Verify policy claims from commerce protocols (UCP, ACS, ACP)
- bluefox_check_safety— Classify content against the 12 GARM brand-safety categories
- bluefox_preflight_check— Run a pre-flight safety check on an agent response before delivery
- bluefox_preflight_batch— Run pre-flight safety checks on multiple responses in one call
- bluefox_egress_gate— Scan an outbound payload with the fail-closed egress gate before it leaves your fleet
- edge_pre_transaction_verify— Run a full trust check (brand, price, policy, compliance) before a purchase
- bluefox_get_brand— Retrieve the raw brand definition record for a brand slug
- bluefox_get_brand_profile— Get a comprehensive, agent-ready brand profile
- bluefox_brand_health— Recorded health snapshot, where one has been computed, from how the AI models the collection observed represented a brand — no snapshot has been computed from the relit collection, so no model figure is stated; each served snapshot carries its own models-observed count and as-of date (collection paused June 4 – September 14, 2026; relit September 14, 2026 on a minimal neutral scope; figures are dated as of collection and have not been recomputed since June 4, 2026 unless dated otherwise; relit scope 8 seed brands × 37 prompts × 5 models, one tick per day, from 2026-09-14T00:49:39Z)
- bluefox_brand_narratives— Top claims and themes AI models associate with a brand
- bluefox_brand_models— Per-model sentiment breakdown across AI models
- bluefox_brand_recommendations— How often AI models recommend a brand across product categories
- bluefox_competitive_radar— Compare a brand's health and recommendation rate against competitors
- bluefox_brand_insights— AI-generated strategic insights and recommendations for a brand
- bluefox_narrative_analyze— Score how a piece of text frames a brand; each dimension reports whether it could be scored
- bluefox_brand_drift— Recorded drift in AI-model perception of a brand at a 7d or 30d horizon
- edge_brand_analytics_summary— Summary of how AI agents interacted with a brand over the last 30 days
- edge_brand_analytics_trends— Week-over-week query trends for a brand
- bluefox_register_agent— Register an AI agent with Edge to build a verification reputation
- bluefox_track_claim— Register one or more claims to monitor for accuracy across AI models
- bluefox_subscribe_monitor— Create a webhook subscription for brand monitoring alerts
- edge_check_reputation— Check your agent's behavioral reputation score on Edge
- edge_create_watcher— Create a custom watcher that delivers structured events when a brand changes
- edge_list_watchers— List your custom watchers, filterable by brand, status, or enabled state
- edge_get_watcher_events— Get triggered events for a watcher — poll as a backup or alternative to webhooks
- edge_advisor_recommend— Recommend the best MCP server for your requirements from behavioral data
- edge_advisor_compare— Compare 2–5 MCP servers side by side across observation dimensions
- edge_advisor_categories— List available MCP server categories with entity counts
- edge_ops_latest— Get the latest operational-intelligence events from a feed
- edge_ops_check— Get all P1 (critical) ops events across feeds from the last 24 hours
- edge_ops_subscribe— Subscribe to ops-feed webhooks for event notifications
Rate Limits & Errors
Usage is metered as a daily call quota per API key. The free tier allows 100 API calls per UTC day; paid plans (starter, pro, enterprise) have higher daily quotas — the plans are on the pricing page. (The REST /verify endpoints additionally cap claim-verification calls at 10 per UTC day; that inner meter never binds MCP tool calls.) When the daily quota is exhausted, MCP tool calls return a JSON-RPC error, code -32000 DAILY_CAP_EXCEEDED — not HTTP 429 — until the quota resets at midnight UTC. Separately from the call quota, the free tier includes 1 signed mint per UTC day (resets 00:00 UTC); over-cap mints return HTTP 402 with x402 payment instructions rather than 429. See pricing for plan details.
Errors arrive in two grammars on this surface. Transport- and auth-layer failures from the gateway are returned as application/problem+json; failures inside an MCP tool call are returned as JSON-RPC 2.0 error objects. Handle both when integrating.
Example Tool Call
When an agent calls bluefox_check_brand, the MCP client sends:
{
"tool": "bluefox_check_brand",
"arguments": {
"brand": "Nike",
"claims": ["Offers free returns within 60 days for members"],
"depth": "auto"
}
}And receives:
{
"brand": "nike",
"results": [
{
"verdict": "accurate",
"confidence": 0.95,
"evidence": "Nike allows free returns within 60 days for Nike Members."
}
],
"brand_definition_version": 1,
"verified_at": "2026-08-05T00:00:00Z",
"latency_ms": 142
}