ANIME INTELLIGENCE / AI PURCHASE INTELLIGENCE

ANIME INTELLIGENCE / AI PURCHASE INTELLIGENCE

Version 3.10.2

Production MCP endpoint:

https://anime-intelligence.goodmy0312.workers.dev/mcp

Production API base:

https://anime-intelligence.goodmy0312.workers.dev

Repository:

io.github.Mike05102/anime-intelligence-mcp

Overview

ANIME INTELLIGENCE / AI PURCHASE INTELLIGENCE is an MCP + API + x402 commerce-intelligence service designed for autonomous AI agents.

The service currently has three main capability groups:

1. Anime / collectible purchase intelligence
2. PARTS INTELLIGENCE for industrial, replacement and discontinued parts
3. Generic commerce search for larger product markets

Travel adapters also exist in the codebase, but live travel inventory remains disabled until provider credentials are configured.

The service is built on:

• Cloudflare Workers
• Supabase
• MCP
• OpenAPI
• x402
• Solana USDC
• Yahoo Shopping integration
• eBay Browse integration
• Rakuten affiliate routing
• public merchant / purchase pages

The design goal is to let an AI agent move from:

search → identify → evaluate → decide → select seller → pay for intelligence → hand off to purchase

without requiring a human to manually assemble information from multiple sources.

1. Anime / Collectible Intelligence

The original ANIME INTELLIGENCE engine focuses on physical Japanese anime collectibles.

Supported product categories include:

• scale figures
• Nendoroid
• figma
• prize figures
• plush
• acrylic goods
• keychains
• badges
• model kits
• trading cards
• collaboration apparel
• collaboration sneakers
• other physical anime merchandise

The system is designed to handle difficult identity problems such as:

• multiple editions of the same character
• rereleases
• manufacturer differences
• Japanese-only product names
• JAN/EAN differences
• prize vs retail versions
• seller-title noise
• incomplete marketplace listings
• ambiguous multilingual search queries

The service resolves a canonical product before returning higher-value purchase intelligence.

Free Anime Search

GET /v1/search

Example:

/v1/search?query=Hatsune%20Miku

Use this for candidate discovery when the exact product is not yet known.

2. Paid Anime x402 Intelligence

Paid endpoints use Solana USDC via x402.

Current paid endpoints:

The service resolves identity before payment when possible so an agent does not pay for the wrong product.

x402 Payment Flow

1. Agent calls a paid endpoint without payment proof.
2. Server returns HTTP 402.
3. Agent reads the PAYMENT-REQUIRED response header.
4. Agent creates the required Solana USDC payment.
5. Agent retries the exact same URL with PAYMENT-SIGNATURE.
6. X-PAYMENT is also accepted as a compatibility alias.
7. After verification and settlement, the endpoint returns HTTP 200.
8. Successful responses include PAYMENT-RESPONSE.

Current x402 network:

• Solana mainnet
• USDC
• x402 v2

3. PARTS INTELLIGENCE

Version 3.10.x adds a second specialized information business:

PARTS INTELLIGENCE

It targets industrial, replacement, maintenance, discontinued and difficult-to-source parts.

Example categories:

• PLC
• sensors
• relays
• switches
• servo motors
• servo amplifiers
• inverters
• power supplies
• pneumatic components
• cylinders
• valves
• bearings
• motors
• connectors
• industrial control components
• machine replacement parts
• appliance / service parts
• camera parts
• audio parts
• tool parts
• IT / server replacement parts

The commercial value comes from resolving questions such as:

• What exact part is this?
• Is it discontinued?
• What is the successor?
• Is there a verified replacement?
• Is a candidate truly compatible?
• Which official document supports the relationship?
• Is the original still available?
• What is the best sourcing route?

PARTS Data Model

PARTS INTELLIGENCE uses Supabase as the system of record.

Core tables:

parts_catalog

Canonical part identity.

Typical fields:

• manufacturer
• brand
• part number
• normalized part number
• name
• category
• lifecycle status
• discontinued date
• successor part number
• specifications
• dimensions
• electrical attributes
• mechanical attributes
• official URL
• source confidence

parts_relations

Explicit relationships between parts.

Supported relation types:

• successor
• substitute
• compatible
• incompatible
• used_in

This table is one of the most important assets in the service.

parts_documents

Evidence and official reference documents.

Examples:

• official PDF
• catalog
• manual
• datasheet
• specification sheet
• replacement table
• compatibility note

parts_offers

Observed sourcing information.

Examples:

• merchant
• price
• currency
• condition
• stock status
• URL
• observation time

parts_sources

Configured official manufacturer sources.

The initial source registry contains:

• OMRON
• Mitsubishi Electric
• SMC
• IDEC
• Panasonic Industry
• Yaskawa
• Oriental Motor
• THK
• NSK
• Azbil

PARTS Compatibility Policy

PARTS INTELLIGENCE does not infer compatibility merely because:

• product names look similar
• dimensions look similar
• model numbers look similar
• marketplace sellers claim interchangeability

A positive substitute / compatibility recommendation requires an explicit relation record with supporting evidence.

Unverified candidates are returned separately.

This is a deliberate safety and data-quality rule.

PARTS Endpoints

GET /v1/parts/status

Returns PARTS database and ingestion status.

GET /v1/parts/search

Free part search.

Example:

/v1/parts/search?part_number=E2E-X5ME1

or:

/v1/parts/search?query=OMRON%20sensor

GET /v1/parts-intelligence

Paid full PARTS intelligence.

Price:

0.05 USDC

Typical output includes:

• canonical part identity
• lifecycle state
• discontinuation status
• successor relationship
• verified substitute relationships
• unverified candidate relationships
• incompatibility evidence
• official documents
• sourcing observations
• purchase / sourcing decision

4. PARTS Ingestion Mode in v3.10.2

Version 3.10.1 included a larger autonomous manufacturer-site crawler.

Version 3.10.2 intentionally removes that crawler from the Cloudflare Worker runtime to reduce deployment size and runtime complexity.

Current ingestion mode:

Supabase bulk import + admin upsert

This keeps the production Worker smaller and more stable.

Initial target:

• several thousand high-quality parts
• then approximately 10,000 high-value FA / industrial parts
• then expand by manufacturer and category

No new external vendor API key is required for the PARTS core.

Existing Yahoo / eBay integrations may optionally enrich sourcing information.

PARTS Database Setup

Run:

PARTS_INTELLIGENCE_v3_10_2_SCHEMA.sql

once in the existing Supabase SQL Editor.

The schema creates:

• parts_catalog
• parts_relations
• parts_documents
• parts_offers
• parts_sources
• parts_ingest_runs

It also seeds the initial official manufacturer source registry.

5. Generic Commerce Search

The server also exposes a more general AI purchase-intelligence layer.

POST /v1/commerce/search

Supported domains currently include:

• consumer_electronics
• automotive_parts
• enterprise_procurement
• mro_and_industrial_components
• electronic_components

Current provider integrations reuse existing configured commerce sources.

Identity Guards

Different domains use different recommendation rules.

Consumer electronics

Generic market search is allowed.

Automotive parts

An exact OE / MPN / part number is required before recommendation.

Enterprise procurement / MRO / electronic components

An exact manufacturer part number is required before recommendation.

The service deliberately avoids recommending technically incompatible components from broad keyword similarity alone.

6. Travel Adapters

Travel code remains present in the Worker.

Current architecture supports:

• Booking.com Demand
• Duffel
• Expedia Rapid
• Amadeus adapter slot

Current production status:

awaiting_credentials

No travel provider is considered live unless:

GET /v1/travel/status

reports the provider as configured.

Travel is not currently the focus of the service because those providers require external API credentials.

7. MCP

MCP endpoint:

https://anime-intelligence.goodmy0312.workers.dev/mcp

The server supports Streamable HTTP MCP.

Relevant tools include:

• search_anime_product
• paid anime intelligence tools
• record_purchase_intent
• commerce_search
• parts_search
• parts_intelligence
• parts_status
• travel tools

MCP metadata is designed so AI agents can understand:

• when to use a tool
• why the tool is paid
• expected output
• x402 payment behavior
• identity requirements
• compatibility restrictions

8. OpenAPI and Discovery

OpenAPI:

/openapi.json

LLM-readable service description:

/llms.txt

x402 discovery metadata:

/.well-known/x402

MCP discovery metadata:

/.well-known/mcp.json

Agent profile:

/agent/profile

Agent services:

/agent/services

Health:

/health

9. Public Purchase Pages

Rakuten affiliate URLs are not directly redistributed through API responses.

Instead, the API can return a public product page hosted by ANIME INTELLIGENCE.

Example pattern:

/shop/product/<canonical-product-id>

The public page can contain merchant routing and affiliate links in a human-readable media context.

This separation reduces ambiguity between machine-facing intelligence and merchant-facing affiliate routing.

10. Telemetry and KPI

The service records conversion telemetry in Supabase.

Important x402 funnel events include:

• API request
• canonical product selection
• x402 gate entered
• payment challenge issued
• payment retry received
• payment attempt
• payment verified
• paid call
• settlement success
• settlement failure

Important commerce events include:

• purchase intent created
• merchant selected
• purchase URL served
• affiliate click
• checkout started
• purchase confirmed

External Revenue KPI

The system separates:

• true external commercial traffic
• crawlers / monitors
• admin tests
• known self-tests
• contextless synthetic probes

Important target KPIs include:

confirmed_external_shopping_intent_calls > 0
external_canonical_product_selected > 0
payment_retry_received > 0
external_payment_attempts > 0
external_payment_verified > 0
external_paid_calls > 0
external_revenue_usdc > 0

A self-payment from the operator's own wallet should not be treated as external commercial revenue.

11. Health Check

Production health endpoint:

https://anime-intelligence.goodmy0312.workers.dev/health

A healthy deployment should currently report:

{
  "ok": true,
  "version": "3.10.2",
  "supabase": "ok"
}

PARTS status should report:

active_supabase_core

and:

new_external_api_keys_required: false

12. Deployment

Main production file:

ANIME_INTELLIGENCE_v3_10_2_FULL_WORKER.txt

Deployment workflow:

1. Open the Cloudflare Worker editor.
2. Replace the entire existing Worker with the new Worker file.
3. Deploy.
4. Verify /health.
5. Confirm version = 3.10.2.
6. Copy the same Worker source to GitHub worker.js.
7. Replace server.json.
8. Replace README.
9. Confirm the MCP Registry workflow succeeds.

13. Supabase

The project uses the existing Supabase environment and credentials.

PARTS INTELLIGENCE does not require an additional database account.

The Worker reuses the current Supabase service-role / secret configuration.

No anonymous database write policy is required.

14. Current Production Status

As of v3.10.2:

• Cloudflare Worker: active
• Supabase: active
• Anime intelligence: active
• x402: active
• MCP: active
• OpenAPI: active
• PARTS INTELLIGENCE core: active
• PARTS external API-key requirement: none
• Generic commerce search: active where existing provider credentials are configured
• Travel: adapter code present, live providers not configured
• Autonomous PARTS crawling: disabled in Worker
• PARTS bulk population: Supabase import / admin upsert

15. Current Priority

The next growth phase is to build the PARTS database itself.

Priority order:

1. populate several thousand canonical industrial parts
2. normalize manufacturer + part-number identity
3. attach official documents
4. add lifecycle / discontinued status
5. add verified successor relationships
6. add verified substitute / compatibility relationships
7. add sourcing observations
8. grow toward ~10,000 high-value records
9. expose more of the resulting intelligence through MCP + x402

The long-term asset is not merely the number of parts.

The highest-value data is the evidence-backed relationship graph:

part → successor
part → verified substitute
part → incompatible alternative
part → machine / equipment usage
part → official document
part → current sourcing option

That relationship graph is what turns PARTS INTELLIGENCE from a search index into paid decision intelligence.