aimlapi

aimlapi

Connect your coding agent to AI/ML API with one command.

npx aimlapi claude

The first run signs you in through the browser (or a device code over SSH), creates an API key, points the agent at AI/ML API and starts it. Add --dry-run to see what would change without touching anything, and --undo to revert it.

Let your agent set it up

Tell your coding agent (Claude Code, Codex, OpenCode, Cline, OpenClaude, OpenClaw, Hermes Agent, zero or ZCode):

Read https://docs.aimlapi.com/quickstart/cli/aimlapi-setup.md and follow it to connect this agent to AI/ML API.

It signs you in through the browser (you approve; you never paste a key into the chat), shows what it will change with --dry-run, applies it after you confirm, and tells you how to undo it.

In Claude Code you can also install it as a skill and then just ask it to "set up AI/ML API":

mkdir -p ~/.claude/skills/aimlapi-setup
curl -fsSL https://raw.githubusercontent.com/aimlapi/api-docs/main/skills/aimlapi-setup/SKILL.md \
  -o ~/.claude/skills/aimlapi-setup/SKILL.md

Supported agents

Agent	Command
Claude Code	npx aimlapi claude
Codex	npx aimlapi codex
OpenCode	npx aimlapi opencode
Cline	npx aimlapi cline
OpenClaude	npx aimlapi openclaude
OpenClaw	npx aimlapi openclaw
zero	npx aimlapi zero
ZCode (Z.ai, GLM models)	npx aimlapi zcode
Hermes Agent	npx aimlapi hermes

The agent itself must already be installed; if it is not, aimlapi prints the install command and exits.

OpenClaw has no per-run provider switch: aimlapi openclaw writes the AI/ML API provider and default model into its openclaw.json (a running Gateway reloads it) and then starts openclaw. That config points at aimlapi's credentials file instead of holding the key, so sign in with aimlapi login (setting AIMLAPI_API_KEY alone is not enough) and run openclaw secrets reload after signing in with a new key. ZCode's terminal agent (zcode) has no published package yet: build it from its repository. aimlapi zcode --config also sets up the ZCode desktop app, which shares the provider config with zcode (and gets its own copy when its settings keep its data in another folder). A provider you add inside an aimlapi zcode session is saved into aimlapi's per-session file, not ZCode's own, and is gone after aimlapi zcode --undo or aimlapi logout: add your providers in a plain zcode session or the desktop app.

Commands

aimlapi <agent> [--model <id>] [--config] [--dry-run] [--undo] [-- <agent args>]
aimlapi login [--device]   # sign in explicitly
aimlapi logout             # revoke the CLI's API key
aimlapi status             # balance, key limit and usage
aimlapi models             # models available for an agent
aimlapi update             # update aimlapi to the latest version
aimlapi --version

For a permanent aimlapi command install it globally:

npm install -g aimlapi

aimlapi-cli is the same package under another name (npx aimlapi-cli claude, npm install -g aimlapi-cli); it installs the same aimlapi command. Install only one of the two globally.

Platforms

macOS and Linux on x64 and arm64. Windows (x64, arm64) is experimental.

This package is a thin Node.js launcher (Node.js >= 18). The aimlapi binary ships in a platform-specific optional dependency (@ai-ml.api/cli-<os>-<arch>) that npm installs automatically for your OS and CPU; there are no install scripts. If you see "platform package is not installed", reinstall without --omit=optional / --no-optional.

Links

• AI/ML API: https://aimlapi.com
• Changelog: CHANGELOG.md