Build with AI

This documentation is built to be consumed by AI coding agents (Claude Code, Cursor, Windsurf, and others) as well as humans. If you're building an app on the Bagisto API — or extending the API itself — with an AI assistant, point it at the resources below.

llms.txt — the machine-readable index

A single file lists every documented endpoint with a one-line description and a link, so an agent can discover the whole API surface in one fetch:

• Index: /llms.txt — every Shop and Admin endpoint, grouped by transport and menu.
• Full content: /llms-full.txt — the complete documentation concatenated into one file for full ingestion.

These follow the llms.txt convention. An agent should fetch llms.txt first to map the surface, then open the specific endpoint page for the exact request and response shape.

Agent skills

The bagisto/api-agent-skills repository packages this knowledge as installable agent skills — reusable instructions that teach an agent how to install the stack, build apps on the API, and extend the API package.

Install all skills into your agent:

npx skills add bagisto/api-agent-skills

Install for a specific agent, or a single skill:

npx skills add bagisto/api-agent-skills -a claude-code
npx skills add bagisto/api-agent-skills --skill "bagisto-api-build-app"

Skill	Use it to
bagisto-api-install	Install Bagisto + the API package
bagisto-api-build-app	Build a storefront / admin / mobile / custom UI on the API
bagisto-api-develop	Add or extend an endpoint in the API package
bagisto-api-dev-cycle	Run the package's tests, lint, and cache cycle
bagisto-api-code-review	Review a change to the package
bagisto-api-git	Branch / stage / prepare a PR

Optional: local docs MCP server

The skills repo also ships an optional local MCP server that lets an agent search this documentation on demand — no hosting required. It indexes the API pages locally and exposes search_api_docs, list_endpoints, and get_doc tools.

claude mcp add bagisto-api -- node /absolute/path/to/api-agent-skills/mcp/src/index.mjs

You don't need the MCP server to build with AI — llms.txt and llms-full.txt cover the same need statically. See the mcp/README.md in the skills repo for setup.

How an agent should use these docs

1. Fetch /llms.txt to discover the endpoint you need.
2. Open that endpoint's page for the exact request body, response shape, and errors — never guess a payload.
3. For end-to-end flows, follow the Recipes — ordered walkthroughs that chain real endpoint calls.
4. Mind the GraphQL rules: select result fields (cartId, orderId, success) on action mutations — not id; inputs are camelCase; the Shop and Admin GraphQL endpoints are separate.