New: URL → MCP via Cloudflare Browser Rendering

Bring your data.
Get a ready-to-use MCP.

Upload a CSV, SQL dump, or paste any URL. ShipMCP provisions an isolated Postgres database, auto-generates MCP tools from your schema, and serves them on the Cloudflare edge — in under a minute.

check_circle No credit card check_circle 500 MB free check_circle 60-second deploy
claude_desktop_config.json
// drop this into Claude Desktop's config
{
  "mcpServers": {
    "products": {
      "url": "https://shipmcp.sh/products/mcp",
      "headers": {
        "Authorization": "Bearer smcp_..."
      }
    }
  }
}
check_circle 7 tools available · 1,247 rows p99 · 84ms
bolt

60s deploy

Upload to live MCP

lock_person

Per-tenant DB

Physical isolation

Engineered for scale

Everything between a database and an LLM.

ShipMCP handles the unglamorous middle: provisioning, isolation, schema introspection, tool generation, edge routing, agent discovery. You bring the data.

database

Unified Data Ingestion

Upload structured files or paste a URL and let Cloudflare Browser Rendering extract markdown or JSON. Schemas auto-detected, columns auto-typed, foreign keys auto-discovered.

CSV JSON · NDJSON SQL Dump Excel URL → Markdown URL → JSON URL → Crawl
lock_person

Per-Tenant Isolation

Dedicated Neon Postgres project per endpoint. Bearer token auth with rate limiting. Not RLS — physical separation.

<100ms

p99 query latency

Edge-Native

Cloudflare Workers + Hyperdrive pooling across 300+ POPs.

build

Auto-Generated MCP Tools

Schema introspection produces 5–8 well-named tools per table — search, filter, get-by-id, count-by, plus FK-aware cross-table joins. Zero handwritten boilerplate.

No connector code Foreign keys detected
list_products
get_products_by_id
search_products
filter_products
count_products_by_category
get_orders_for_product
+ 1 more
3-step setup

From file to live MCP in under 60 seconds.

1

Upload or extract

Drop a CSV, JSON, SQL dump, or Excel file. Or paste a URL — ShipMCP uses Cloudflare Browser Rendering to convert pages to markdown, structured JSON, or crawl whole sites.

2

We provision

A dedicated Neon Postgres project spins up. Data loads, schema is introspected, MCP tools are generated, llms.txt is published. All async, all on the edge.

3

Connect any agent

Copy the endpoint URL and bearer token into Claude Desktop, Cursor, or any MCP-compatible agent. Agents fetch llms.txt for context, then call your tools.

https://shipmcp.sh/products/llms.txt
# ShipMCP — `products`

> MCP endpoint exposing products,
> categories, orders from products.csv.

## Tools

- list_products — paginated listing
- search_products — full-text search
- filter_products — by category,
  price range, in_stock, created_at

## Schema

### `products` — ~1,247 rows
  - id (int4, PK)
  - name (text)
  - price (numeric)
  - category (text)
check_circle Public · agent-discoverable /llms.txt
Three vendors. Three bills.

A deliberately lean stack.

No Stripe integration. No Auth0. No DataDog. Cloudflare for everything edge, Neon for per-tenant Postgres, Polar as Merchant of Record for billing.

Cloudflare

Pages · Workers · D1 · R2 · Queues · Hyperdrive · Browser Rendering

Neon

Per-tenant Postgres projects, isolated by design.

Polar

Merchant of Record · global VAT/GST · metered billing.

Ready to ship your data?

Free tier: 500 MB storage, 1,000 MCP calls/month. No credit card required.

Up and running in 2 minutes.