Documentation

Quickstart

Go from data to live MCP endpoint in three steps. Everything below is real and current as of today.

Sign up & upload

Create a free account, then either drop a structured data file or paste a URL. Supported file formats:

CSV JSON NDJSON SQL Dump (.sql) Excel (.xlsx, .xls)

Or for URL extraction, pick a mode:

→ Markdown — fetches one page, converts to clean markdown, parses tags/links/headings into a document corpus.
→ Crawl — same-origin BFS up to N pages, all stored as markdown documents.
→ JSON — uses Cloudflare's structured-extraction endpoint to coerce the page into records.

Wait for provisioning

ShipMCP runs the following pipeline asynchronously (typically 30–60 seconds):

Creates a dedicated Neon Postgres project
Loads your data (file parser or web extractor)
Introspects the schema
Generates 5+ MCP tools per table (search, filter, get-by-id, count-by, FK joins)
Publishes llms.txt at /{slug}/llms.txt
Sends you a "live" email and updates the dashboard

Connect Claude Desktop (or any MCP client)

From the endpoint detail page, copy the URL and create a token. Then add to claude_desktop_config.json:

~/Library/Application Support/Claude/claude_desktop_config.json

{
  "mcpServers": {
    "products": {
      "url": "https://shipmcp.sh/products-abc12/mcp",
      "headers": {
        "Authorization": "Bearer smcp_..."
      }
    }
  }
}

Restart Claude Desktop. The agent calls initialize, reads the instructions field (which points at the public llms.txt), then calls tools/list to see what's available.

Agent discovery via llms.txt

Every active endpoint serves an llms.txt at two well-known paths (no auth required):

https://shipmcp.sh/{slug}/llms.txt https://shipmcp.sh/{slug}/.well-known/llms.txt

The file describes the endpoint's purpose, full schema, all available tools with argument signatures, FK relationships, and connection instructions. It's auto-generated from your live database every time you re-ingest.

Tool generation rules

For each table in your database, ShipMCP produces tools based on column types:

Column trait	Generated tool
text, varchar	`search_` (full-text), `filter_` (exact match)
int, numeric	`filter_*` (min/max range), sort
timestamptz, date	`filter_*` (date range), sort by recency
boolean	`filter_*` toggle
primary key	`get_*_by_id`
foreign key	`get__for_` (cross-table join)
low-cardinality text	`count__by_` (aggregation)

Ready to ship?

Free tier is enough to evaluate end-to-end.

Start free →