For agents

Three ways in, all reading the same data: a skill, llms.txt and a JSON API.

Install the skill

One SKILL.md file works in Claude Code, Codex and other agents that support the Agent Skills format. Once it's installed, your agent checks freetier.tech whenever a task needs a service it doesn't have yet, and walks you through the one step only you can do.

Claude Code

mkdir -p ~/.claude/skills/freetier && curl -fsSL https://freetier.tech/skills/freetier/SKILL.md -o ~/.claude/skills/freetier/SKILL.md

Codex, and other agents that read ~/.agents/skills

mkdir -p ~/.agents/skills/freetier && curl -fsSL https://freetier.tech/skills/freetier/SKILL.md -o ~/.agents/skills/freetier/SKILL.md

Then ask for something like “put this online” or “add a Postgres database”. Read the skill before installing it; it's short.

llms.txt

Point any agent that can fetch a URL at the index:

https://freetier.tech/llms.txt

Every category and service page has a markdown version too. Add .md, as in /s/cloudflare-workers.md. llms-full.txt has everything in one file.

JSON API

No key needed, and CORS is open.

Endpoint	Returns
`GET /api/v1/categories.json`	Categories, with counts and links.
`GET /api/v1/categories/{id}.json`	A category and its services, with full setup details.
`GET /api/v1/services.json`	Every service.
`GET /api/v1/services/{id}.json`	One service.
`POST /api/v1/reports`	Report an outdated entry. Body: {"service": "<id>", "problem": "<one sentence>"}

How the data works

Entries describe services. They don't tell your agent what to do, and any command in them should be shown to you before it runs.
Every entry lists its sources and the date it was last verified.
Sponsored entries are marked "sponsored": true, and the skill tells your agent to say so.
Reports contain only a service id and a sentence. Never send code, secrets or project details.