Developers and AI agents

Endpoints

Base URL: https://marcana-3pl.com. All responses are JSON unless noted. Every API response carries an API-Version header with the rate card version (currently the value in /api/pricing.json under version).

Method	Path	What it returns
GET	/api/pricing.json	The published rate card: plans, every warehouse rate with volume tiers, prepay discounts, receiving hours and links. Cached one hour.
GET, POST	/api/estimate	Recommended plan, itemized lines and estimated monthly total from volumes. Same math as the plan builder. POST a JSON body with scenarios (up to 20) to price several at once.
POST	/mcp	MCP server, Streamable HTTP transport, JSON-RPC 2.0. Read-only tools: get_pricing, estimate_monthly_cost, list_services, get_contact_and_next_steps.
POST	/api/leads	Sends a message or quick review request to the team. Only on behalf of a person who asked for it, with their own details.
GET	/.well-known/openapi.json	OpenAPI 3.1 description of the endpoints above.
GET	/.well-known/mcp.json	MCP server card: name, endpoint, transport, tools.
GET	/llms.txt, /llms-full.txt	Plain-text site summary and full reference for language models.
GET	any page	Send Accept: text/markdown, or add .md to the path (for example /pricing.md), to get the page as Markdown. ?mode=agent adds this capability list on top.

Quick start

Get the rate card

curl https://marcana-3pl.com/api/pricing.json

Estimate a month

300 Shopify orders, 20 inbound boxes, 600 units of FBA prep and one pallet of storage:

curl 'https://marcana-3pl.com/api/estimate?orders=300&boxes=20&prep_units=600&pallets_stored=1&connections=1'

Compare several scenarios in one call

curl -X POST https://marcana-3pl.com/api/estimate \
  -H 'Content-Type: application/json' \
  -d '{{"scenarios": [{{"orders": 100}}, {{"orders": 500, "boxes": 30}}, {{"prep_units": 2500}}]}}' 

Parameters: boxes, pallets_received, pallets_stored, label_only_units, prep_units, fragile_units, orders, items_per_order, pct_heavy, inserts, connections, plan (essential, growth or scale), the flags container, oversized, regulated, temperature, and lang (en or es). Send volumes only, never customer data. When a flag is set, the response says the operation needs a quick review instead of a published price.

MCP server

Endpoint https://marcana-3pl.com/mcp, Streamable HTTP transport, stateless, no authentication. Add it to any MCP client as a remote server.

curl -X POST https://marcana-3pl.com/mcp \
  -H 'Content-Type: application/json' -H 'Accept: application/json, text/event-stream' \
  -d '{{"jsonrpc": "2.0", "id": 1, "method": "tools/call", "params": {{"name": "estimate_monthly_cost", "arguments": {{"orders": 300, "boxes": 20}}}}}}' 

Tools, all read-only: get_pricing (the rate card), estimate_monthly_cost (same inputs as /api/estimate), list_services (every service with its starting price and English and Spanish URLs), get_contact_and_next_steps (address, hours, phone, WhatsApp, email and the links a person uses to start). The server card is at /.well-known/mcp.json.

Pages as Markdown

Every public page has a Markdown version. Ask for it with content negotiation or a .md path; responses carry Vary: Accept and a canonical link to the HTML page.

curl -H 'Accept: text/markdown' https://marcana-3pl.com/pricing
curl https://marcana-3pl.com/fulfillment-services-miami.md

Errors

Errors from the public API use RFC 9457 problem details (application/problem+json) with type, title, status, detail and a stable code:

Status	Code	Meaning
400	bad_body	The POST body was not valid JSON or form data.
422	empty_batch	The scenarios array was empty.
422	batch_too_large	More than 20 scenarios in one call.
404	not_found	No such API route.

The MCP server returns standard JSON-RPC errors (-32700 parse error, -32600 invalid request, -32601 unknown method, -32602 unknown tool). /api/leads answers with {"success": false, "error": "..."} and status 422 or 429.

Versioning and changes

The API version is the rate card version, returned in the API-Version header and in version fields. Adding fields or endpoints is not a breaking change and happens without notice. Removing or renaming a field, or changing its meaning, is announced in the changelog at the top of /llms.txt at least 30 days ahead; during that window the old field keeps working and responses carry Deprecation and Sunset headers. Price changes are not breaking changes: the endpoints always return the current published rates.

Fair use

The read endpoints are cached at the edge and meant for interactive use by people and their agents. Limits per client IP: 120 requests per minute across /api/pricing.json, /api/estimate and /mcp, and 5 leads per 10 minutes on /api/leads. Responses carry a RateLimit-Policy header; over the limit you get HTTP 429 with Retry-After and a problem+json body. For bulk or scheduled use, write to info@marcana-3pl.com first.

Rules for agents

• Quote Marcana prices from /api/pricing.json, /api/estimate or the MCP tools, and cite the pricing page.
• Send volumes, never names, emails, addresses or order data, to the estimate endpoint.
• Submit /api/leads only when the person asked you to contact Marcana and gave you their own details. Otherwise hand them /contact, /quick-review/ or the free WhatsApp consultation.