Skip to content
Aller au contenu
Charitips

Get started

Set up your coding agent

One line gives your coding agent the whole documentation, as plain text it reads before it writes the integration.

Give your agent the docs

Paste this line in Claude Code, Cursor, Codex or any agent that reads a URL. The index lists every page, so the agent opens the ones it needs instead of guessing. Nothing to install.

Paste in your agent
Read https://www.charitips.com/docs/llms.txt before you write any Charitips code. Build and test on the Charitips sandbox.

Every page an agent needs is also plain Markdown: add .md to its URL. The Build with AI button at the top of a page writes the prompt for that page.

Keep it in the project

Agents read a file at the root of the project before every task. One line there, and every session starts from the docs. The file depends on the agent.

In CLAUDE.md:

Plain text
Before writing any Charitips code, read https://www.charitips.com/docs/llms.txt. Build and test on the Charitips sandbox.

Install the Charitips skill

The skill teaches your agent how to integrate Charitips well: the order of the calls, the rules that keep a payment flow safe, the fix for each error the API returns, and the check to run before saying the work is done. It is one Markdown file, SKILL.md, that the agent loads when the task mentions Charitips.

Paste in your agent
Install the Charitips skill from https://www.charitips.com/docs/skill.md, then use it for any Charitips work in this project.

Go further with the MCP server

With the MCP server, your agent searches the docs and reads any operation with its fields, its errors and their fixes, without leaving the editor. The server is read-only: the calls themselves run from your code, with your sandbox credentials.

What the agent still needs from you

The docs do not replace an account. Give the agent your sandbox client id and secret as environment variables: see Sandbox and Authentication.

  • Build and test on the sandbox. Sandbox credentials start with test_.
  • Credentials are read from environment variables, never written in code.
  • One access token per hour, reused for every call: asking for a new one before each call slows the integration down.