AI agents and LLMs
Give your coding agent the real contract instead of a guess. Every page here is available as plain Markdown, and one skill teaches an agent how to integrate Akedly from start to finish.
Machine-readable docs
The site publishes the same pages in forms an agent can read without scraping HTML. They are generated from the pages themselves each time the docs are built, so they never disagree with what you read here.
| Address | What it holds |
|---|---|
/llms.txt | The docs index in the llms.txt format: a short summary, the rules that hold for every method, and a link and one-line description for each page. |
/llms-full.txt | Every current page in one file. It is large, so prefer a single page when you can. |
/<page>.md | One page as Markdown: add .md to the page address, as in /quickstart.md. The home page is /index.md. |
/errors.json | The error codes from the docs tables as JSON, to generate a typed error handler. |
/sitemap.xml | Every page that is open to search engines. |
The error list in /errors.json is the common cases, not a closed set. Handle a code you do not recognise by its HTTP status, as the errors page explains.
The Akedly integration skill
The skill is a folder in the open Agent Skills format. An agent that supports skills loads it when you ask for an Akedly integration. It then:
- picks between the V2.0 Widget and the V1.2 REST API, and adds passkeys on top ("OTP once, Passkey forever");
- sends one-way utility messages over WhatsApp and SMS from templates, with signed requests and webhook handling;
- keeps the API key, the pipeline ID and the widget secret on your server;
- verifies webhook signatures on the raw body;
- generates error handling from the error catalogue;
- tests with dev mode before it reports the work as done.
It carries a snapshot of the core pages, so it also works offline. When the agent has network access it reads the live pages first.
| Item | Value |
|---|---|
| Version | 1.1.0 |
| Download | akedly-integration-skill.zip |
| SHA-256 | akedly-integration-skill.zip.sha256, rebuilt with every docs release |
| Docs snapshot | The docs-version field in the skill's SKILL.md names the docs build it was made from. |
| License | Apache-2.0 |
Install the skill
Pick your agent. Each agent reads skills from its own folders, and those conventions change, so check your agent's documentation if a path here does not work.
# Option 1: the plugin marketplace (Claude Code 2.1.224 or later)
/plugin marketplace add https://docs.akedly.io/claude-plugins/marketplace.json
/plugin install akedly@akedly
# Option 2: unzip into your personal skills folder
curl -fsSLO https://docs.akedly.io/downloads/akedly-integration-skill.zip
unzip akedly-integration-skill.zip -d ~/.claude/skills/
Start a new session after installing, then ask your agent for the integration, for example "Add Akedly phone verification to this Express app."
Check the download
Compare the checksum before you install, especially if you fetch the file in a script.
curl -fsSLO https://docs.akedly.io/downloads/akedly-integration-skill.zip
curl -fsSLO https://docs.akedly.io/downloads/akedly-integration-skill.zip.sha256
shasum -a 256 -c akedly-integration-skill.zip.sha256
The zip holds text files only: Markdown, JSON, a licence and a version stamp. It has no scripts, so installing it runs nothing.
Keep keys out of the conversation
Never paste a key or a secret into an agent
Your API key, pipeline ID, widget secret and webhook signing secret belong in environment variables on your server. Tell the agent the variable names and set the values yourself. The skill never asks for a secret, and an agent that asks for one should be stopped. If you pasted one by mistake, replace it in the dashboard.
Without installing anything
Give an agent the address https://docs.akedly.io/llms.txt and ask it to read the pages it needs, or paste one page as Markdown: add .md to its address, as in /quickstart.md. Every page header also links here, with a "Build with AI agents" link. All of it uses the same page text as the Markdown files above. The per-platform "Prompt your LLM" pages are retired: the skill and the Markdown pages above replace them, and an old link to one lands here.
Where to go next
V2.0 Widgets
—the hosted widget, with the strongest protection tier
V1.2 REST API
—build your own screen on the REST API
Passkeys on V2
—let returning users skip the code
Webhooks
—how signed results reach your server
Errors
—the failure envelope and how to branch on a code