Developers
The agent surface, as it exists.
Shipeezi publishes its public knowledge and read API for agents. Everything on this page is live on this site, and none of it needs a credential. This is the human-readable map of the machine-readable files.
Start here
The machine-readable files, and what they are for.
Everything here is live on this site. Nothing is protected: no key, no signup and no token. This page is a map of the files an agent or an integration can read.
openapi.json
The public read API in OpenAPI 3.1: shipments, events and schema, with an operationId per operation. The same document is at /openapi.yaml.
MCP manifest
The registry-format server manifest at /.well-known/mcp.json, pointing at the MCP endpoint at /mcp over Streamable HTTP.
auth.md
How to authenticate, which is: you do not. No key, no signup, no token. It follows the auth.md structure so an agent that expects those sections finds them.
Page index
/api/v1/pages returns every published page with its title, description and markdown URL. Add a slug to read one page.
Site search
/api/v1/search?q=sscc returns ranked full-text search across the site, with a snippet for each hit.
Markdown alternates
Every page has a .md twin at the same path, and any page URL returns markdown for Accept: text/markdown. Responses carry Vary: Accept.
The public read API
Read the trace spine, no credentials.
The API exposes the trace Shipeezi owns as a DCSA-shaped track and trace surface. Every operation is a read. Send no Authorization header; it is ignored.
| Method and path | What it returns |
|---|---|
GET /api/public/v1/shipments/{reference} | One shipment and its leg chain, by reference. |
GET /api/public/v1/events | Track and trace events across shipments. |
GET /api/public/v1/schema | The event and reference schema used by the API. |
GET /api/v1/pages | The site's page index, with markdown URLs. |
GET /api/v1/pages/{slug} | One page, as JSON or markdown. |
GET /api/v1/search?q={query} | Ranked full-text search across the site. |
The API base for the trace endpoints is api.shipeezi.com; the full contract is in openapi.json.
MCP tools
The MCP server exposes four tools.
| Tool | What it does |
|---|---|
search_shipeezi_site | Full-text search over every published page. |
get_shipeezi_page | One page as markdown, navigation and footer removed. |
list_shipeezi_pages | Every published page with its title, description and markdown URL. |
get_shipeezi_pricing | The pricing page as markdown. |
Point an agent at llms.txt to start, or connect the MCP server at /mcp.
Conventions
What to expect from the surface.
Reads only
Every public operation is a read. There is no write surface exposed here.
Errors are JSON
An error carries an error string and a repeated status. You will never see a 401 or a 403 from this API.
Rate limits and backoff
The JSON API has a generous per-IP ceiling. On a 429, read the RateLimit-* headers and Retry-After, then back off.
Crawling is welcome
See robots.txt and the sitemap. Nonexistent paths return a real 404, never a 200 with an app shell.
Correlation
Send an optional X-Request-Id on any request and it is echoed on the response.
The tenant platform is separate
The authenticated platform behind app.shipeezi.com is a different system, not documented here. Access to it is a commercial conversation.
Build on it.
Point your agent at llms.txt, then call the read API. If you need the authenticated platform, that is a commercial conversation.