AI Agents

Everything on this site is available to an AI agent in a form it can consume directly: a hosted MCP server, an installable skill, and the whole documentation set as Markdown and OpenAPI. This page lists the exact commands and URLs.

Fastest paths

Connect the MCP server

Agents that speak the Model Context Protocol should connect to the server rather than call REST by hand. It exposes headless commerce reads and order creation, branded gift-card PDF templates, and order-page configuration, publishes the field catalog at runtime, and ships prompts that encode the correct workflow. It shows an agent only the tools its key holds the scope for.

bash

claude mcp add --transport http lifepeaks https://api.lifepeaks.dk/mcp \
  --header "Authorization: Bearer lp_live_<your-key>"

For Cursor, VS Code, and every other client, see MCP Server — Install.

Install the order-page skill

The lifepeaks-order-page skill teaches an agent to read and customize a Lifepeaks order page. Install it from this site with the skills CLI:

bash

npx skills add https://docs.lifepeaks.dk/skills/lifepeaks-order-page/SKILL.md

To load it manually — into a system prompt, a repository, or a client that does not use the CLI — fetch the raw files:

Skill definition — https://docs.lifepeaks.dk/skills/lifepeaks-order-page/SKILL.md
Field catalog reference — https://docs.lifepeaks.dk/skills/lifepeaks-order-page/references/fields.md

Machine-readable surfaces

Documentation index — https://docs.lifepeaks.dk/llms.txt

Every page with a one-line description, following the llms.txt convention. Use this when the agent can fetch pages on demand and should decide for itself what to read.

Whole corpus — https://docs.lifepeaks.dk/llms-full.txt

Every page inlined into a single Markdown document. Use this when the agent cannot fetch again and the entire surface has to fit in one pass.

One page as Markdown — https://docs.lifepeaks.dk/raw/<page>.md

The Markdown source of a single page, for example https://docs.lifepeaks.dk/raw/endpoints.md. Every page on this site has one; the introduction is at /raw/index.md.

OpenAPI 3.1, v2 — https://docs.lifepeaks.dk/openapi/v2.yaml

The authoritative description of every v2 endpoint. Prefer it over prose whenever exact field names, types, or response shapes matter.

OpenAPI 3.0, v1 (legacy) — https://docs.lifepeaks.dk/openapi/v1.json

The same for the legacy OAuth2 API, for agents working against an existing v1 integration.

Every documentation page also carries a Copy page control next to its title, which copies that page's Markdown or hands it to Claude or ChatGPT directly.

What the skill covers

Authentication and the scope each operation needs
The draft-review-publish workflow, from orderpage_list_fields through orderpage_publish_draft
The full writable-field catalog with types, maximum lengths, and per-locale support
Type validation rules for string, bool, color, csv, gtm, and int
Writing translations by drafting one locale at a time
What is deliberately not writable, and the errors returned when it is attempted
Worked end-to-end examples

Notes for agent implementers

The writable field catalog is dynamic. Call GET /v2/order-page/schema or orderpage_list_fields before writing, rather than hardcoding field names.
Prefer a draft over patching the live page. Nothing reaches buyers until it is published, and a mistaken publish can be rolled back by restoring an earlier revision.
Every draft edit and publish needs the newest ETag. A stale one is rejected, which is what stops two editors overwriting each other.
Non-catalog fields are rejected with 422 Unprocessable Entity, not silently ignored.
Write tools take an idempotency_key, and most also require confirm: true. Send confirm only after the user has approved that specific side effect.
The key returned by POST /v2/api-keys is shown exactly once. An agent that stores credentials must store it on receipt.
Minimum scopes for the order page: orderpage:read to read, orderpage:write to draft and publish.