If you build with a coding agent, you can hand it these two files instead of pasting documentation into a prompt. Both are plain text, public, and need no account.

llms.txt

A machine-readable index of every page here, with one line on what each covers. Follows the llmstxt.org convention.

skills.md

A self-contained integration skill: the working request shapes, the error table, and the four mistakes that cost real money.

​
Claude Code

Save the skill into your project and Claude picks it up automatically whenever a task involves sending SMS:
Or, for a one-off, just point it at the file:

​
Codex, Cursor, and other agents

Add the URL to whatever file the tool reads for project context — AGENTS.md, .cursorrules, or similar:

​
What the skill already knows

It carries the things that are expensive to learn by trial, so an agent gets them right the first time rather than after a bill:
  • Each envelope is charged separately. The same text to 500 recipients costs at least 500 credits — there is no shared-body discount.
  • One non-GSM character re-prices the whole message, from 160 characters per part to 70. A 140-character message is 1 credit; add one accented letter and it is 3.
  • callbackUrl is accepted and ignored. An agent that builds a webhook listener for it will wait forever; delivery comes from polling.
  • 202 means queued, not delivered.
Give the agent a token from a test account with a small balance the first time. A send that is wrong in a loop is charged like any other send.

​
Keeping credentials out of it

The skill tells the agent to read the token from an environment variable rather than write it into code. Set it yourself:
Never paste a token into a prompt, a repository, or an agent’s configuration file. A token carries full account permissions, including the ability to spend your credits. See Authentication.