Let your agent set this up

A skill is a short instruction file a coding agent reads when a task matches it. This one tells the agent how to get an Inference APIs key without a browser, which config file to write for the tool it is running in, and how to check that it worked. It is the same steps a person would follow, written for the agent.

https://inferenceapis.com/skill/SKILL.md

Install

mkdir -p ~/.claude/skills/inferenceapis-setup && curl -fsSL https://inferenceapis.com/skill/SKILL.md -o ~/.claude/skills/inferenceapis-setup/SKILL.md
# then, in Claude Code: "set up Inference APIs as my provider"

mkdir -p ~/.codex/skills/inferenceapis-setup && curl -fsSL https://inferenceapis.com/skill/SKILL.md -o ~/.codex/skills/inferenceapis-setup/SKILL.md
# then: codex "set up Inference APIs as my provider, my email is you@example.com"

curl -fsSL https://inferenceapis.com/skill/SKILL.md   # paste it into the agent's instructions, or point the agent at the URL

What the agent does

1. Calls POST /v1/signup with the email you give it and receives an API key. A confirmation email goes to you; opening it adds $1 of starter credit.
2. Reads GET /v1/models, which lists prices, context length and capabilities, and picks a model for the job.
3. Writes the config file for the tool it is running in: Claude Code, Codex, opencode, Aider, Continue, VS Code, or the OpenAI and Anthropic SDKs.
4. Sends one request and reports the result. If it fails, every error carries a docs_url that explains the code.

Two things stay with you: opening the confirmation email, and adding funds on the billing page. An agent cannot pay.

Tested on 2026-09-18 with codex-cli 0.154.0: given only an email address, it created the account, wrote ~/.codex/config.toml and confirmed the key with a request. Machine-readable overview for agents without a skill: /llms.txt.