Getting_Started

This system mirrors the production backend: account wallet funding, agent-scoped API keys, layered policies, and purchase tracking through MCP or direct API.

Hosted_MCP

Stateless MCP implementation. Each request carries an agent API key forwarded to the backend.

json
{
  "mcpServers": {
    "ag3n7-store": {
      "transport": {
        "type": "streamable-http",
        "url": "https://ag3n7.store/mcp",
        "headers": {
          "Authorization": "Bearer ask_live_xxxxx.yyyyy"
        }
      }
    }
  }
}

Stdio_MCP

Standard I/O implementation for clients that prefer local process spawning.

json
{
  "mcpServers": {
    "ag3n7-store": {
      "command": "node",
      "args": ["/absolute/path/to/agent_store_backend/dist/mcp/server.js"],
      "env": {
        "AGENT_STORE_API_BASE_URL": "https://ag3n7.store/v1/agent",
        "AGENT_STORE_AGENT_API_KEY": "ask_live_xxxxx.yyyyy"
      }
    }
  }
}

Purchase_Flow

Recommended sequence: narrow the catalog, inspect checkout requirements for custom products, quote with final configuration, then purchase.

text
1. catalog_categories
2. catalog_search { category, q, limit }
3. catalog_checkout_requirements { catalogItemId }
4. quote_create { catalogItemId, quantity, configuration? }
5. purchase_create { quoteId | (catalogItemId, quantity, configuration?), idempotencyKey }
6. purchase_get_status { purchaseId }
7. service_list / service_get
8. domain_* or server_* management actions
9. router_api_key_create for OpenAI-compatible credit services
Open_Domain_FlowOpen_Server_FlowOpen_LLM_Router_FlowOpen_Router_PricingOpen_Phone_API_FlowOpen_Storage_Flow

LLM_Router

Buy credits with the purchase API, then mint a separate router key for an OpenAI-compatible base URL. The router key is distinct from the agent purchase key.

text
1. catalog_search { category: "api_tokens", q: "router credits", limit: 1 }
2. catalog_checkout_requirements { catalogItemId }
3. quote_create { catalogItemId, quantity: 1, configuration: { creditAmountUsd: 25 } }
4. purchase_create { quoteId, idempotencyKey }
5. service_get { serviceId }
6. router_api_key_create { serviceId, label: "production" }
7. GET https://ag3n7.store/openai/v1/models
8. POST https://ag3n7.store/openai/v1/chat/completions
python
from openai import OpenAI
import os
 
client = OpenAI(
    api_key=os.environ["AGENT_STORE_ROUTER_API_KEY"],
    base_url="https://ag3n7.store/openai/v1",
)
 
response = client.chat.completions.create(
    model="claude-sonnet-4",
    messages=[
        {"role": "system", "content": "Be concise."},
        {"role": "user", "content": "Summarize yesterday's deploy."},
    ],
)
 
print(response.choices[0].message.content)
Open_Full_FlowBrowse_Provider_Prices

Phone_API

Rent phone numbers across multiple countries, then manage top-ups, inbound SMS, and missed calls through agent tools after provisioning.

text
1. catalog_search { category: "telephony", q: "phone number", limit: 3 }
2. catalog_checkout_requirements { catalogItemId }
3. quote_create {
     catalogItemId,
     quantity: 1,
     configuration: {
       countryCode: "US",
       numberType: "local",
       capabilities: ["sms", "voice"],
       locality: "new-york"
     }
   }
4. purchase_create { quoteId, idempotencyKey }
5. service_get { serviceId }
6. phone_balance_topup { serviceId, amountUsd: 10 }
7. phone_message_list / phone_message_get
8. phone_missed_call_list
Open_Phone_Flow

Storage

Object storage, block volumes, and snapshot retention follow the same quote-first managed-service pattern as other custom products.

text
1. catalog_search { category: "storage", q: "object" }
2. catalog_checkout_requirements { catalogItemId }
3. quote_create {
     catalogItemId,
     quantity: 1,
     configuration: {
       region: "us-east",
       storageClass: "object",
       capacityGb: 250
     }
   }
4. purchase_create { quoteId, idempotencyKey }
5. service_get { serviceId }
6. storage_bucket_credentials_create / storage_volume_attach
7. storage_snapshot_restore
Open_Storage_Flow

Agent_API

Direct API exposing full management capabilities including wallet, history, custom checkout, domains, and VPS actions. The dashboard checkout uses the same quote and purchase payloads through an internal account-scoped checkout actor.

TypeScript
typescript
const apiKey = process.env.AGENT_STORE_AGENT_API_KEY;
 
const categories = await fetch('https://ag3n7.store/v1/agent/catalog/categories', {
  headers: { Authorization: `Bearer ${apiKey}` },
}).then((res) => res.json());
 
const items = await fetch(
  'https://ag3n7.store/v1/agent/catalog/items?category=domains&q=ai&limit=5',
  { headers: { Authorization: `Bearer ${apiKey}` } },
).then((res) => res.json());
 
const quote = await fetch('https://ag3n7.store/v1/agent/quotes', {
  method: 'POST',
  headers: {
    Authorization: `Bearer ${apiKey}`,
    'content-type': 'application/json',
  },
  body: JSON.stringify({
    catalogItemId: '<catalog-item-uuid>',
    quantity: 1,
    configuration: {
      domainName: 'launchpad.com',
      periodYears: 1,
      autoRenew: true,
      privacyProtection: true,
    },
  }),
}).then((res) => res.json());
 
const purchase = await fetch('https://ag3n7.store/v1/agent/purchases', {
  method: 'POST',
  headers: {
    Authorization: `Bearer ${apiKey}`,
    'content-type': 'application/json',
  },
  body: JSON.stringify({
    quoteId: quote.quote.id,
    idempotencyKey: 'deploy-domain-1',
  }),
}).then((res) => res.json());
cURL
bash
API_BASE_URL="https://ag3n7.store/v1/agent"
API_KEY="ask_live_xxxxx.yyyyy"
&nbsp;
curl -X POST "$API_BASE_URL/quotes"   -H "Authorization: Bearer $API_KEY"   -H "Content-Type: application/json"   -d '{
    "catalogItemId": "<catalog-item-uuid>",
    "quantity": 1,
    "configuration": {
      "domainName": "launchpad.com",
      "periodYears": 1,
      "autoRenew": true,
      "privacyProtection": true
    }
  }'
&nbsp;
curl -X POST "$API_BASE_URL/purchases"   -H "Authorization: Bearer $API_KEY"   -H "Content-Type: application/json"   -d '{
    "quoteId": "<quote-id>",
    "idempotencyKey": "deploy-domain-1"
  }'
Python
python
import os
import requests
&nbsp;
api_base_url = "https://ag3n7.store/v1/agent"
api_key = os.environ["AGENT_STORE_AGENT_API_KEY"]
headers = {
    "Authorization": f"Bearer {api_key}",
    "Content-Type": "application/json",
}
&nbsp;
quote_response = requests.post(
    f"{api_base_url}/quotes",
    headers=headers,
    json={
        "catalogItemId": "<catalog-item-uuid>",
        "quantity": 1,
        "configuration": {
            "domainName": "launchpad.com",
            "periodYears": 1,
            "autoRenew": True,
            "privacyProtection": True,
        },
    },
    timeout=30,
)
quote_response.raise_for_status()
quote = quote_response.json()
&nbsp;
purchase_response = requests.post(
    f"{api_base_url}/purchases",
    headers=headers,
    json={
        "quoteId": quote["quote"]["id"],
        "idempotencyKey": "deploy-domain-1",
    },
    timeout=30,
)
purchase_response.raise_for_status()
purchase = purchase_response.json()

Endpoints: GET /v1/agent/catalog/categories, GET /v1/agent/catalog/items, GET /v1/agent/catalog/items/:catalogItemId/checkout, GET /v1/agent/wallet, GET /v1/agent/policies/effective, POST /v1/agent/quotes, POST /v1/agent/purchases, GET /v1/agent/purchases, GET /v1/agent/purchases/:purchaseId, GET /v1/agent/services, GET /v1/agent/services/:serviceId, domain endpoints under /v1/agent/services/:serviceId/domain, and server endpoints under /v1/agent/services/:serviceId/server.

Security_Model

SECURITY_PROTOCOL Keep agent keys server-side, use stable idempotency keys for retries, and treat approval states as authoritative.

Current_Contract: Production_AlphaAccess_Control_Surface