Skip to content

Get started

Build with AI

How to give Claude Code, Cursor or ChatGPT the right context from the Messajo docs, with llms.txt, per-page Markdown and OpenAPI.

A coding assistant gets the API right more often when it reads the real documentation instead of guessing. Messajo publishes the docs in formats made for that.

What Where What for
Docs index https://docs.messajo.com/llms.txt Short index: API description and links to the full docs, the condensed docs and OpenAPI. To read a single page, use its .md
Full docs https://docs.messajo.com/llms-full.txt Everything in a single file, for a large context
Short docs https://docs.messajo.com/llms-small.txt Abridged version, for a small context
One page <page url>.md The page as plain Markdown
OpenAPI /openapi.json The exact contract of every route, body and error

The llms*.txt files are generated from the Portuguese pages only. The per-page .md files exist in every language.

Append .md to the address of any page:

https://docs.messajo.com/start/quickstart.md
https://docs.messajo.com/en/start/quickstart.md
https://docs.messajo.com/es/start/quickstart.md

Next to every page title there is a Copy page control. It copies the page Markdown to the clipboard and its menu offers View as Markdown, Open in ChatGPT and Open in Claude. The last two open the assistant with a request to read that page.

The API contract is available as JSON, one file per language:

The interactive reference is generated from the same file, and so are the types of the TypeScript SDK. Hand over the OpenAPI when the assistant needs the exact shape of a request body.

The path is the same in Claude Code, Cursor and other coding assistants:

  1. In the conversation. Paste the address of llms.txt or of the page’s .md and ask it to read it before writing code.
  2. In the project’s instruction file. Write the address once in the file your assistant reads at the start of every session (such as CLAUDE.md or the project rules), and every session starts with the docs at hand.
  3. For exact formats. Give it openapi.json, or the .md of the page for the route in question, instead of asking the assistant to remember.

Example snippet for the instruction file:

Messajo API (official WhatsApp, Instagram and Messenger):
- Docs index: https://docs.messajo.com/llms.txt
- API contract: https://docs.messajo.com/openapi.json
- Before writing code that calls the API, read the route's .md page.
- Use the @messajo/sdk SDK; never invent fields outside the OpenAPI.

Every API error has the same body, with three fields that help whoever is debugging (see Error format):

  • code: a stable identifier, safe to handle in code.
  • docs_url: a link to the code in Errors, with its meaning and what to do.
  • request_id: the same value as the X-Request-Id header (see X-Request-Id for support).

Paste the whole error body into the assistant: the docs_url leads to the right explanation.

A good first request, with llms.txt already in the conversation:

Read https://docs.messajo.com/en/send/text.md and https://docs.messajo.com/en/webhooks/verify-signature.md.
Then, with @messajo/sdk:
1. send a text to 5511999999999 using the key in MESSAJO_API_KEY;
2. create an endpoint that receives webhooks and verifies the Messajo-Signature
header before handling the event.
If a field is not in the docs, ask instead of making it up.

Supporting pages: Send text and Verify the signature.